WO2022062304A1

WO2022062304A1 - Method and device for deploying image recognition service on container cloud

Info

Publication number: WO2022062304A1
Application number: PCT/CN2021/075492
Authority: WO
Inventors: 肖乐源
Original assignee: 亮风台（上海）信息科技有限公司
Priority date: 2020-09-25
Filing date: 2021-02-05
Publication date: 2022-03-31
Also published as: CN111930473B; CN111930473A

Abstract

Provided is a method for deploying an image recognition service on a container cloud. The method comprises: after a basic service in an image recognition service is deployed, acquiring a deployment file corresponding to an image recognition business service in the image recognition service, wherein business services in the image recognition business service comprise at least one of the following: an image recognition management service, a front-end resource service, an index service, at least one recognition service, a feature extraction service, an account service, a content management service, a publishing service, an image recognition API service and a resource storage service; and for each business service in the image recognition business service, creating a container set on a container cloud according to the deployment file, starting the business service, loading a corresponding configuration file by means of the container set according to a configuration service address corresponding to the business service, and managing the business service by means of a container management platform.

Description

Method and device for deploying image recognition service on container cloud

This application is based on the application with the CN application number of 202011021515.1 and the filing date of 2020.09.25, and claims its priority. The disclosure of the CN application is hereby incorporated into this application as a whole.

technical field

The present application relates to the field of communications, and in particular, to a technology for deploying an image recognition service on a container cloud.

Background technique

In the prior art, image recognition services are usually deployed on physical servers or virtual machines (VMs), requiring manual operations to increase hardware server resources, install an operating system, and modify the configuration to start the service, adjust the capacity and load of the cluster, and sometimes even The service cluster needs to be restarted, and additional log and error collection methods need to be built to handle service restarts and error records. When facing the concurrent processing of large-capacity image loading, the processing capacity is not easy to expand.

SUMMARY OF THE INVENTION

An object of the present application is to provide a method and device for deploying an image recognition service on a container cloud.

According to an aspect of the present application, a method for deploying an image recognition service on a container cloud is provided, the method comprising:

After the basic service in the image recognition service is deployed, a deployment file corresponding to the image recognition business service in the image recognition service is obtained, wherein the deployment file includes a service corresponding to each business service in the image recognition business service Name, image file address, number of container sets, configuration service address, the business services in the image recognition business service include at least one of the following: image recognition management service, front-end resource service, indexing service, at least one identification service, feature extraction service , account service, content management service, publishing service, image recognition API service, resource storage service;

For each business service in the image recognition business service, create a container set on the container cloud according to the deployment file and start the business service, and load the corresponding configuration file according to the configuration service address corresponding to the business service through the container set , and manage the business service through the container management platform, wherein the configuration file contains the mapping relationship between the service names and service addresses of other business services that the business service depends on, so that the business service can be The mapping relationship calls other business services that the business service depends on;

Wherein, the image recognition business service deployed through the container cloud performs at least any one of the following:

Receive an image publishing service request for a first identification map through the publishing service, obtain the first identification map, and provide the first identification map to the indexing service; obtain the first identification map through the indexing service identification map, generate a first index file corresponding to the first identification map, and return the first storage address corresponding to the first index file to the publishing service; obtain the first storage address through the publishing service, providing the first storage address to a first identification service corresponding to a first atlas corresponding to the first identification map in the at least one identification service; obtaining the first storage through the first identification service address, load the first index file according to the first storage address; or,

Receive an image recognition service request for a picture to be recognized sent by a target application through the image recognition API service, and obtain a user atlas corresponding to the to-be-recognized picture; use the image recognition API service from the database in the basic service The service obtains the second identification service corresponding to the user atlas in the at least one identification service, and provides the to-be-identified picture and the second identification service to the feature extraction service; obtains the feature extraction service through the feature extraction service. the picture to be identified and the second identification service, extract the feature information of the picture to be identified, and provide the feature information to the second identification service; obtain the feature information through the second identification service, Match the feature information with the index file currently loaded by the second identification service; if the matching is successful, obtain a second identification corresponding to the second index file that matches the feature information through the second identification service map information, and return the second recognition map information to the image recognition API service, obtain the second recognition map information through the image recognition API service, and obtain the second recognition map information from the basic service. the second AR material, and return the second AR material to the target application.

According to another aspect of the present application, a method for publishing images based on an image recognition service deployed on a container cloud is provided, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and managed through the container The platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: a front-end resource service, an indexing service, at least one identification service, a content management service, and a publishing service. Identify each business service in the business service, create a container set on the container cloud according to the deployment file, and start the business service. The container set loads the configuration file corresponding to the business service, and the configuration file contains the dependencies of the business service. The mapping relationship between service names and service addresses of other business services, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Among them, the method includes:

Receive an image publishing service request for a target recognition map through the publishing service, obtain the target recognition map, and provide the target recognition map to the indexing service according to the service address of the indexing service;

Obtain the target identification map through the indexing service, generate a target index file corresponding to the target identification map, and return the target storage address corresponding to the target index file to the publishing service;

Obtain the target storage address through the publishing service, and provide the target storage address to a target recognition service corresponding to the target atlas set corresponding to the target recognition map in the at least one recognition service;

Obtain the target storage address through the target identification service, and load the target index file according to the target storage address.

According to another aspect of the present application, an image recognition method based on an image recognition service deployed on a container cloud is provided, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and managed through the container The platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, and image recognition API service. a business service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the services of other business services that the business service depends on The mapping relationship between the name and the service address, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Among them, the method includes:

Receive, through the image recognition API service, an image recognition service request sent by the target application for the image to be recognized, and obtain a user atlas corresponding to the image to be recognized;

The target recognition service corresponding to the user atlas in the at least one recognition service is obtained from the database service in the basic service corresponding to the image recognition service through the image recognition API service, and the target recognition service of the service is extracted according to the feature. The service address provides the image to be recognized and the target recognition service to the feature extraction service;

Obtain the picture to be recognized and the target recognition service through the feature extraction service, extract the feature information of the picture to be recognized, and provide the feature information to the target recognition service according to the service address of the target recognition service Serve;

Obtain the feature information through the target recognition service, and match the feature information with the index file currently loaded by the target recognition service;

If the matching is successful, obtain the target recognition map information corresponding to the target index file matching the feature information through the target recognition service, and return the target recognition map information to the image recognition API service. The API service obtains the target recognition map information, obtains the target AR material corresponding to the target recognition map information from the basic service, and returns the target AR material to the target application.

According to another aspect of the present application, a network device for deploying an image recognition service on a container cloud is provided, the device comprising:

One module is used to obtain the deployment file corresponding to the image recognition business service in the image recognition service after the basic service in the image recognition service is deployed, wherein the deployment file contains each image recognition service in the image recognition service. The service name, image file address, number of container sets, and configuration service address corresponding to each business service. The business services in the image recognition business service include at least one of the following: image recognition management service, front-end resource service, indexing service, at least one Recognition service, feature extraction service, account service, content management service, publishing service, image recognition API service, resource storage service;

The first and second modules are used to create a container set on the container cloud according to the deployment file for each business service in the image recognition business service and start the business service, and configure the service according to the corresponding business service through the container set. The corresponding configuration file is loaded from the address, and the business service is managed through the container management platform. The business service invokes other business services on which the business service depends according to the mapping relationship;

According to another aspect of the present application, a network device for image publishing based on an image recognition service deployed on a container cloud is provided, wherein an image recognition service corresponding to the image recognition service is deployed on the container cloud, and the container cloud is The management platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: a front-end resource service, an indexing service, at least one identification service, a content management service, and a publishing service. For each business service in the image recognition business service, a container set is created on the container cloud according to the deployment file and the business service is started. The container set loads the configuration file corresponding to the business service, and the configuration file contains the dependencies of the business service. The mapping relationship between the service names and service addresses of other business services, so that the business service calls other business services that the business service depends on according to the mapping relationship;

Among them, the equipment includes:

The second module is configured to receive an image publishing service request for a target recognition map through the publishing service, obtain the target recognition map, and provide the target recognition map to the indexing service according to the service address of the indexing service. ;

The second module is configured to obtain the target identification map through the indexing service, generate a target index file corresponding to the target identification map, and return the target storage address corresponding to the target index file to the publishing service;

The second and third modules are configured to obtain the target storage address through the publishing service, and provide the target storage address to the target recognition service corresponding to the target atlas set corresponding to the target recognition map in the at least one recognition service ;

The second and fourth modules are configured to obtain the target storage address through the target identification service, and load the target index file according to the target storage address.

According to another aspect of the present application, a network device for image recognition based on an image recognition service deployed on a container cloud is provided, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and an image recognition service corresponding to the image recognition service is deployed on the container cloud. The management platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, and image recognition API service. For each business service, a container set is created on the container cloud according to the deployment file and the business service is started. The container set loads the configuration file corresponding to the business service, and the configuration file contains the configuration files of other business services that the business service depends on. The mapping relationship between the service name and the service address, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Among them, the equipment includes:

The 31 module is configured to receive, through the image recognition API service, an image recognition service request sent by a target application for a picture to be recognized, and obtain a user atlas corresponding to the to-be-recognized picture;

The third and second modules are configured to obtain the target recognition service corresponding to the user atlas in the at least one recognition service from the database service in the basic service corresponding to the image recognition service through the image recognition API service, and according to The service address of the feature extraction service provides the to-be-recognized picture and the target recognition service to the feature extraction service;

The third and third modules are used to obtain the picture to be recognized and the target recognition service through the feature extraction service, extract the feature information of the picture to be recognized, and extract the feature information according to the service address of the target recognition service. to provide said target identification service;

The third and fourth modules are used to obtain the feature information through the target recognition service, and match the feature information with the index file currently loaded by the target recognition service;

The three and five modules are used to obtain the target recognition map information corresponding to the target index file matching the feature information through the target recognition service if the matching is successful, and return the target recognition map information to the image recognition API service , obtain the target recognition map information through the image recognition API service, obtain the target AR material corresponding to the target recognition map information from the basic service, and return the target AR material to the target application.

According to an aspect of the present application, a device for deploying an image recognition service on a container cloud is provided, wherein the device includes:

processor; and

memory arranged to store computer-executable instructions which, when executed, cause the processor to:

According to another aspect of the present application, an image publishing device based on an image recognition service deployed on a container cloud is provided, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and managed through the container The platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: a front-end resource service, an indexing service, at least one identification service, a content management service, and a publishing service. Identify each business service in the business service, create a container set on the container cloud according to the deployment file, and start the business service. The container set loads the configuration file corresponding to the business service, and the configuration file contains the dependencies of the business service. The mapping relationship between service names and service addresses of other business services, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Among them, the equipment includes:

processor; and

According to another aspect of the present application, an image recognition device based on an image recognition service deployed on a container cloud is provided, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and managed through the container The platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, and image recognition API service. a business service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the services of other business services that the business service depends on The mapping relationship between the name and the service address, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Among them, the equipment includes:

processor; and

a memory arranged to store computer-executable instructions which, when executed, cause the processor to perform the following operations:

According to one aspect of the present application, there is provided a computer-readable medium storing instructions that, when executed, cause a system to:

According to another aspect of the present application, a computer-readable medium for storing instructions is provided, wherein an image recognition service corresponding to an image recognition service is deployed on the container cloud, and the image recognition service is processed through a container management platform. service management, the business services in the image recognition business service include at least one of the following: front-end resource service, indexing service, at least one identification service, content management service, publishing service, for each service in the image recognition business service service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the service name and the service name of other business services that the business service depends on. The mapping relationship between the service addresses, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Wherein, the instruction, when executed, causes the system to perform the following operations:

According to another aspect of the present application, a computer-readable medium for storing instructions is provided, wherein an image recognition service corresponding to an image recognition service is deployed on the container cloud, and the image recognition service is processed through a container management platform. service management, the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, image recognition API service, for each business service in the image recognition business service, according to the deployment file Create a container set on the container cloud and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the service name and service address of other business services that the business service depends on. a mapping relationship, so that the business service invokes other business services on which the business service depends according to the mapping relationship;

Obtain the feature information through the target recognition service, and match the feature information with the index file currently loaded by the target recognition service.

Compared with the prior art, by deploying the image recognition service on the container cloud, the portability and security of the image recognition service can be improved, and the container management platform is used to reduce deployment difficulty, increase deployment flexibility, and accelerate related products. The online speed can also solve the technical problem of rapid expansion of large-scale image recognition services, and realize the dynamic expansion of image recognition services.

Description of drawings

Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

1 shows a flowchart of a method for deploying an image recognition service on a container cloud according to an embodiment of the present application;

Fig. 2 shows the image recognition service architecture diagram of an example of the present application;

3 shows a flowchart of a method for publishing images based on an image recognition service deployed on a container cloud according to an embodiment of the present application;

4 shows a flowchart of an image recognition method based on an image recognition service deployed on a container cloud according to an embodiment of the present application;

FIG. 5 shows a structural diagram of a network device for deploying an image recognition service on a container cloud according to an embodiment of the present application;

FIG. 6 shows a structure diagram of a network device for image publishing based on an image recognition service deployed on a container cloud according to an embodiment of the present application;

FIG. 7 shows a structural diagram of a network device for image recognition based on an image recognition service deployed on a container cloud according to an embodiment of the present application;

FIG. 8 illustrates an exemplary system that may be used to implement various embodiments described in this application.

The same or similar reference numbers in the drawings represent the same or similar parts.

detailed description

The present application will be described in further detail below with reference to the accompanying drawings.

In a typical configuration of the present application, the terminal, the device serving the network, and the trusted party all include one or more processors (for example, a central processing unit (CPU)), an input/output interface, a network interface, and RAM.

Memory may include non-persistent memory in computer-readable media, random access memory (RAM) and/or non-volatile memory, such as read only memory (ROM) or flash memory ( Flash Memory). Memory is an example of a computer-readable medium.

Computer readable media includes both persistent and non-permanent, removable and non-removable media and can be implemented by any method or technology for storage of information. Information may be computer readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, Phase-Change Memory (PCM), Programmable Random Access Memory (PRAM), Static Random-Access Memory (Static Random-Access Memory, SRAM), Dynamic Random Access Memory (DRAM), other types of random access memory (RAM), read only memory (ROM), Electrically-Erasable Programmable Read -Only Memory, EEPROM), flash memory or other memory technology, Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD), or other optical storage , magnetic tape cartridges, magnetic tape-disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

The equipment referred to in this application includes, but is not limited to, user equipment, network equipment, or equipment formed by integrating user equipment and network equipment through a network. The user equipment includes, but is not limited to, any mobile electronic product that can perform human-computer interaction with the user (for example, human-computer interaction through a touchpad), such as a smart phone, a tablet computer, etc., and the mobile electronic product can use any operation. system, such as Android operating system, iOS operating system, etc. Wherein, the network device includes an electronic device that can automatically perform numerical calculation and information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs) ), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Digital Signal Processor (DSP), embedded devices, etc. The network device includes but is not limited to a computer, a network host, a single network server, a plurality of network server sets or a cloud formed by a plurality of servers; here, the cloud is formed by a large number of computers or network servers based on cloud computing, Among them, cloud computing is a kind of distributed computing, a virtual supercomputer composed of a group of loosely coupled computer sets. The network includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, a wireless ad hoc network (Ad Hoc network), and the like. Preferably, the device may also be a program running on the user equipment, network equipment, or a device formed by user equipment and network equipment, network equipment, touch terminal or network equipment and touch terminal integrated through a network.

Of course, those skilled in the art should understand that the above-mentioned devices are only examples, and other existing or possible devices that may appear in the future, if applicable to this application, should also be included within the protection scope of this application, and are included in this application by reference. this.

In the description of this application, "plurality" means two or more, unless expressly and specifically defined otherwise.

FIG. 1 shows a flowchart of a method for deploying an image recognition service on a container cloud according to an embodiment of the present application, and the method includes steps S11 and S12. In step S11, after the basic service in the image recognition service is deployed, a deployment file corresponding to the image recognition business service in the image recognition service is obtained, wherein the deployment file includes each of the image recognition business services. The service name, image file address, number of container sets, and configuration service address corresponding to the business service. The business service in the image recognition business service includes at least one of the following: image recognition management service, front-end resource service, indexing service, at least one recognition service service, feature extraction service, account service, content management service, publishing service, image recognition API service, resource storage service; in step S12, for each business service in the image recognition business service, according to the deployment file in the container Create a container set on the cloud and start the business service, load the corresponding configuration file according to the configuration service address corresponding to the business service through the container set, and manage the business service through the container management platform, wherein the configuration file contains Contains the mapping relationship between the service names and service addresses of other business services that the business service depends on, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, through the container cloud The deployed image recognition business service performs at least any one of the following: receiving an image publishing service request for a first recognition map through the publishing service, obtaining the first recognition map, and providing the first recognition map to The indexing service: obtaining the first identification map through the indexing service, generating a first index file corresponding to the first identification map, and returning the first storage address corresponding to the first index file to the publication service; obtaining the first storage address through the publishing service, and providing the first storage address to a first identification corresponding to a first atlas corresponding to the first identification diagram in the at least one identification service obtain the first storage address through the first identification service, and load the first index file according to the first storage address; or, through the image identification API service, receive the information sent by the target application for the target application Recognize the image recognition service request of the picture, and obtain the user atlas corresponding to the to-be-recognized picture; obtain the corresponding user atlas in the at least one recognition service from the database service in the basic service through the image recognition API service the second identification service, provide the picture to be identified and the second identification service to the feature extraction service; obtain the picture to be identified and the second identification service through the feature extraction service, extract the feature information of the picture to be recognized, and provide the feature information to the second recognition service; obtain the feature information through the second recognition service, and combine the feature information with the second recognition service currently loaded If the matching is successful, obtain the second identification map information corresponding to the second index file matching the feature information through the second identification service, and return the second identification map information to the Image Recognition API service, through the Image Recognition API The service obtains the second identification map information, obtains the second AR material corresponding to the second identification map information from the basic service, and returns the second AR material to the target application.

In some embodiments, containerization technology (docker) is used to support the creation and use of Linux containers. Docker utilizes resource separation mechanisms in the Linux kernel, such as cgroups, and Linux kernel namespaces (namespaces) to create independent containers (containers). ). Containerization is an approach to software development whereby an application or service, its dependencies, and its configuration (abstracted as a deployment manifest file) are packaged together as a container image, where a software container acts as the standard unit of software deployment.

In some embodiments, container management platforms (including but not limited to Kubernetes (K8s), Swam, Meso, etc., where Kubernetes provides a framework for running distributed systems elastically, in the context of Kubernetes as an example for container management platforms described) for deploying, scaling, and managing containerized applications, so groups of hosts running Linux containers can be clustered together, efficiently managed by Kubernetes, and available across public clouds, Private or hybrid cloud deployment hosts.

In some embodiments, Kubernetes consists of a set of nodes, which can be physical hosts or virtual machines. The Kubernetes platform runs on these nodes to form a cluster. A Kubernetes cluster consists of a Master Node (host node) and a worker Node (node) for the actual deployed Pods. A cluster is often composed of multiple Master Nodes. These Master Nodes control all Nodes of the entire cluster, and all task assignments come from the Master Nodes. The Master Node controls the entire cluster, makes global decisions about the cluster (such as scheduling), and detects and responds to cluster events. Node is the real workload node in the K8s cluster. The K8s cluster shares the workload with multiple Nodes, and supports dynamic addition or deletion of Nodes in the cluster. Pod (container set) is assigned to a specific Node for execution, and contains a container group of one or more containers. All containers in the same container set share the same IP address, IPC, hostname, and other resources. Containers are subordinate to Pods. In running applications, libraries and their dependent microservices, containers are the lowest level. By binding an external IP, containers can be accessed by the external network.

In some embodiments, the Kubernetes master node (Master Node) accepts instructions from the outside, and then forwards the instructions to the affiliated Nodes (nodes), this switching work with a large number of services automatically determines the most suitable node for the task, and then The node allocates resources and assigns Pods (container sets) to complete task requests. In some embodiments, each business service in the image recognition service is packaged in a docker image and runs in each Pod in Kubernetes.

In some embodiments, the container cloud platform combines container technology and container management platform in the cloud to provide a platform for building, publishing and running distributed applications, hereinafter referred to as container cloud. In some embodiments, the container cloud runs on multiple sets of servers (for example, x64 servers, x86 servers), deploys multiple business services in the image recognition service on the container cloud, and uses Kubernetes to perform operations on the services deployed on the container cloud. The management of these business services in the image recognition service can solve the technical problem of rapid expansion of large-scale image recognition services, realize the load balancing and dynamic expansion of image recognition services, solve the problem of differences in the deployment of different systems and different hardware environments, and improve the service quality. Portability and security, reduce deployment difficulty, increase deployment flexibility, and accelerate the launch of related products.

In some embodiments, the basic services in the image recognition service include at least one of the following: database, cache database, file storage, configuration, image repository, configuration editor and other related basic services, the basic services can be deployed On the container cloud, it can also be deployed on the periphery of the container cloud (eg, on an external entity cloud server). Preferably, the basic service is deployed on the periphery of the container cloud.

In some embodiments, after the basic service in the image recognition service is deployed, k8s obtains a deployment file corresponding to the image recognition business service in the image recognition service. The deployment file is the basic file for starting the image recognition service, and k8s is specified in it. The required basic content, such as specifying the service name of each business service in the image recognition business service, the image file address, the number of container sets, and the environment variables when the container set is running (such as the address of the configuration service, etc.) in the deployment file, Further, the configuration item key can also be specified in the environment variable, and further, the content of the service (SVC) of each business service (including the exposed port and other content) can also be specified in the deployment file, and k8s is created according to the deployment file. Container set and start each business service in the image recognition business service on the container set. The image recognition business service includes at least one of the following: image recognition management service, front-end resource service, indexing service, recognition service, feature extraction service, account service , content management service, publishing service, image recognition API service, resource storage service, wherein, there may be one or more deployment files, which are not limited here. In some embodiments, each business service corresponds to the same configuration service address, and in other embodiments, each business service corresponds to a different configuration service address, which is not limited herein. There is no restriction on the deployment sequence of different business services here. In some embodiments, the image recognition management service is used to perform service control, such as controlling the allocation of recognition servers, and the like. The front-end resource service is used to manage the back-end or the front-end of the web page, and is also an entry service included in the corresponding client. In some embodiments, divided according to business logic, the image recognition business service includes a second layer service that depends on the basic service and a first layer service that depends on the second layer service, wherein the second layer service The service is used to provide services that are biased towards the underlying algorithm function type (such as at least one service such as indexing service, identification service, feature extraction service, etc.), and has little relevance to specific business logic. The first layer service is used to provide Services that are biased towards upper-level business logic types (such as at least one service such as account service, content management service, publishing service, image recognition API service, etc.) are closely related to specific business logic. First-tier services and second-tier services are divided according to business logic, which does not mean that there is hierarchical division in deployment, but is only used to facilitate understanding of business logic. Dependency means that the relying party directly or indirectly uses or invokes or updates the data, functions, interfaces, etc. of the relying party. For example, the publishing service in the first-tier service needs to call the indexing service in the second-tier service; for example, the first-tier service The content management service in the service updates the database content in the base service, etc. Figure 2 is a service architecture diagram of an example of the application, in which all business services and configuration editors are stored in the mirror warehouse, the mirror address is configured after the service is created in the container cloud, and the corresponding configuration is obtained from the configuration service (etcd) after the service is started. , initialize the mysql database tables and data, and complete the construction of the service architecture; among them, the cluster service entrance is the front-end resource service (webhiar), and the service request is forwarded to the account service (accountsrv), content management service (contentsrv), publishing service (publishsrv), The first layer of services such as image recognition API service (recognizesrv), the first layer depends on indexing services (including cloud indexing service (cloudkey), local indexing service (localkey)), recognition service (searchsrv), feature extraction service (featuresrv) and other third-level services. Second-tier service, the second-tier service relies on basic services such as database service mysql, file storage service fastdfs, and resource storage service (storagesrv). Among them, webhiar is used as the access entry, which is used to deploy the entry service of the API in the management background or the front end of the web page. The access of the application and the API request of the application end are all entrances from this node. When concurrent access occurs, it can be deployed on the server. After building load balancing before, different services in the architecture can expand nodes according to the pressure of different requests. It should be noted that, the various services involved in the above will be described in detail in subsequent embodiments. In some embodiments, by deploying the image recognition service on a container cloud, the portability and security of the image recognition service can be improved.

In some embodiments, for each business service in the image recognition business service, a container is created on the container cloud according to deployment information such as the service name, image file address, number of container sets, etc. corresponding to each business service in the deployment file set and start the business service, the container set loads the configuration file corresponding to the business service from the configuration service according to the configuration service address in the deployment file (for example, the configuration service includes a plurality of configuration files corresponding to different business services respectively , you can find the configuration file corresponding to the business service in the configuration service through the service name of the business service and load it), and use the container management platform (for example, k8s) to identify at least one of the deployed images on the container cloud. Services are managed. In some embodiments, a configuration file needs to be loaded when the business service is started, and the configuration file contains the content required by the business service when it runs, usually including the port number occupied by the business service, the database address of the connection, the address of the cache server, dependencies (need to call ) and other information about the service address of other business services. In some embodiments, K8S creates a container set corresponding to the business service and the number of container sets on the container cloud according to the number of container sets corresponding to the business service in the deployment file, and according to the image file corresponding to the business service in the deployment file address, download the image file corresponding to the business service, start the business service on the container set according to the image file, and then set the environment variable according to the deployment file, and the container set reads the configuration service set in the environment variable. address, load the configuration file corresponding to the business service from the configuration service, and complete the startup of the business service according to the content of the configuration file, and then the business service can process the business request.

In some embodiments, if the basic service is deployed on the container cloud, at least one of the basic service and the image recognition business service deployed on the container cloud can be managed through the container management platform. In other embodiments, if the basic service is deployed on the peripheral service of the container cloud, at least one service among the image recognition business services deployed on the container cloud can be managed through the container management platform. In some embodiments, the k8s master node (Master Node) accepts instructions from the outside, and then forwards the instructions to the attached Node (node), and then allocates resources on the node and assigns Pod (container set) to complete the task request. In some embodiments, each service is packaged in a docker image and runs in each Pod in k8s. In some embodiments, using k8s to manage the image recognition service deployed on the container cloud can solve the technical problem of rapid expansion of large-scale image recognition services, realize load balancing and dynamic expansion of image recognition services, reduce deployment difficulty, and increase Deployment flexibility to accelerate the launch of related products.

In some embodiments, the configuration file includes a mapping relationship between service names and service addresses of other business services that the business service depends on, so that the business service can call other services that the business service depends on according to the mapping relationship. Serve. For example, if the publishing service depends on the indexing service and other business services, and the publishing service needs to call the indexing service and other business services, the configuration file corresponding to the publishing service contains the mapping relationship between the service name of the indexing service and the indexing service address and other services. The mapping relationship between the service name of the service and the addresses of other business services. When the publishing service needs to call the indexing service, the service of the indexing service can be obtained through the mapping relationship between the service name and the service address of the indexing service in the configuration file. address, so that the publishing service can call the indexing service according to the service address of the indexing service.

In some embodiments, the second layer service includes at least one of the following:

1) Indexing service

In some embodiments, the indexing service is a cloud indexing service (cloudkey), or a local indexing service (localkey), or includes both a cloud indexing service and a local indexing service; the cloud indexing service is used to generate cloud-identified images of images. The index file is used to identify the application in the cloud; the local index service is used to generate the index file required for the local identification and used for the local identification application. The index refers to the feature of the extracted picture.

In some embodiments, the indexing service accepts an index identification request, generates an index through a local index and a cloud index, saves the index file to the file storage service, and returns a storage address.

2) Identify the service

The identification service (searchsrv) is used to provide the basic service and core service of the image identification algorithm. In some embodiments, the identification service accepts the identification request, retrieves the index file stored in the identification library of the local machine, and returns identification information (eg, ID) of the successfully matched identification picture.

3) Feature extraction service

The feature extraction service (featuresrv) is a pre-service during recognition, and is used to extract the features of the recognized picture.

In some embodiments, there may be one or more second-layer services, which can be dynamically expanded according to the pressure and traffic conditions requested by the client. It should be noted that the above-mentioned second-layer service is only an example, rather than a limitation of the present application.

In some embodiments, the first-tier service includes at least one of the following:

1) Content Management Services

The content management service (contentsrv) is used to provide relevant management services for identifying gallery, AR material and AR scene. In some embodiments, the content management service includes application management, resource management, atlas management, and AR scene management, and may also include binding the identification picture to the AR material.

2) Publish service

The publishing service (publishsrv) is a management service for publishing pictures to the identification service, and is used to control the publishing of the atlas, and belongs to the core service. In some embodiments, the publishing service manages and allocates existing recognition servers, accepts a request for publishing pictures, calls an index generation module, and allocates recognition servers, so that the recognition server calls a corresponding index loading module to load the index file .

3) Image recognition API service

The image recognition API service (recognizesrv) provides API interface services such as client image recognition, resource acquisition, and local recognition index file download. In some embodiments, the image recognition API service is used to: call a publishing management service to obtain recognition server information; call an image recognition service to retrieve pictures, obtain a recognition interface, query data, and return AR materials corresponding to the images.

4) Account service

The account service is used to provide services related to account management, such as managing user and enterprise-related information. In some embodiments, there may be one or more first-tier services, which may be dynamically expanded according to the pressure and traffic conditions requested by the client. It should be noted that the above-mentioned first-layer service is only an example, rather than a limitation of the present application.

In some embodiments, the basic service includes at least one of the following:

1) Configure the service

The configuration service (etcd) is used to provide the configuration of the corresponding service. In some embodiments, the configuration service is a distributed and consistent KV storage system for shared configuration and service discovery, and the configuration of the corresponding service can be modified and created through the management page of the configuration service. In some embodiments, the service architecture of the present application employs a shared configuration service.

2) Database service

The database service is used to store the data of the system. For example, the database service can use a relational database to store text data in the form of two-dimensional tables, including but not limited to the mysql database. In some embodiments, this application includes, but is not limited to, four databases: an account database, a content database, a publication database, and an identification server database. Among them, the account database saves user and enterprise information, etc.; the content database includes atlas, attribute information of pictures (picture ID, name, atlas, size, type, release status, belonging user, etc.), AR effect binding information, resources Information (video, music, model, etc.), etc.; the publishing database stores the publishing server information and picture publishing information, including picture ID, atlas, index file address, corresponding identification server, etc.; the identification server database generates a A record table, including the ID of the loaded identification map, the address of the index file, the atlas, the release time, etc. In other embodiments, one or more databases may be established to store the above information, which is not limited here.

3) File storage service

The file storage service is used to provide services related to the storage of files (eg, picture files, index files, videos, music, models, and other resource files). In some embodiments, the file storage service includes, but is not limited to, any service related to file storage, such as fastdfs.

4) Cache database service

For example, the cache database service may be redis (Remote Dictionary Server).

5) Mirror warehouse service

The image warehouse service (harbor) is used to provide services of image storage and distribution.

6) Configure the editor service

The configuration editor service (etcdkeeper) is used to provide an application configuration file editing service.

In some embodiments, the base service can run inside the container cloud, or it can run on the periphery of the container cloud. In some embodiments, the base service can be dynamically expanded. It should be noted that the above basic services are only examples, rather than limitations of this application.

In some embodiments, the method further includes deploying a resource storage service (storagesrv) on the container cloud. In some embodiments, the resource storage service is used to provide an intermediate service for file storage, encapsulates a standard storage interface to other services, and can be connected to multiple storage services later. The resource storage service is equivalent to an adapter for adapting to different file storage formats. In some embodiments, the resource storage service needs to be deployed inside the container cloud, and cannot be deployed outside the container cloud.

In some embodiments, for each business service in the image recognition business service, creating a container set on the container cloud according to the deployment file and starting the business service includes: for each business service in the image recognition business service a business service, according to the number of container sets corresponding to the business service in the deployment file, create a container set corresponding to the business service and the number of container sets on the container cloud, and according to the number of container sets corresponding to the business service in the deployment file Image file address, download the image file corresponding to the business service, and start the business service on the container set according to the image file. In some embodiments, during the process of starting the business service by the container set, the container set also needs to load a configuration file to complete the starting of the identification service.

In some embodiments, the image recognition management service is used to manage the at least one recognition service; wherein the method further comprises: for each recognition service of the at least one recognition service, when the recognition service is started , send registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service; receive the registration information through the image recognition management service, and generate the recognition service according to the registration information and store the relevant information of the identification service in the database service in the basic service. In some embodiments, each recognition service will record the domain name information and/or address information of the image recognition management service. After each recognition service is started, it will actively send the image recognition service to the image recognition service according to the domain name information and/or address information The identification management service sends registration information including identification information of the identification service (eg, ID, name, address of the identification service), and informs the image identification management service that the identification service is available, and the image identification management service receives the registration information, and write the registration information into the database service in the basic service, and generate the relevant information of the identification service in the database service (for example, the ID of the currently loaded identification map, index file address, atlas, release time, etc.), In order to record how many recognition services are currently activated and the relevant information of each activated recognition service, the subsequent image recognition management service will regularly monitor the operation of all activated recognition services. When the machine is down, it will search and obtain the relevant information of the downtime identification service from the database service according to the identification information of the downtime identification service.

In some embodiments, the method further includes: for each activated recognition service in the at least one recognition service, periodically sending heartbeat information to the image recognition management service through the activated recognition service; The image recognition management service does not receive the heartbeat information sent by the third recognition service in the at least one recognition service within a predetermined time interval, then the fourth recognition service is determined in the at least one recognition service, and the basic service is determined from the basic service. The database service in obtains the currently loaded atlas information corresponding to the third identification service, so as to load the atlas corresponding to the currently loaded atlas information in the fourth identification service. In some embodiments, each recognition service periodically sends heartbeat information to the image recognition management service, so that the image recognition management service determines whether and when each recognition service fails or is down based on the heartbeat information, eg, each recognition service The service sends heartbeat information to the image recognition management service every predetermined time. If the image recognition management service receives the heartbeat information of a certain recognition service, it can be determined that the recognition service is still running. If the image recognition management service does not receive the heartbeat information after a predetermined time The heartbeat information to an identification service indicates that the identification service is down. In some embodiments, if the image recognition management service does not receive the heartbeat information sent by the third recognition service within a predetermined time interval, it may be determined that the third recognition service is down, and the image recognition management service will retrieve the registered recognition service from the (The registered identification service may include the identification service that is newly started after the third identification service is down, and may also include the identification service that has been started before the third identification service is down.) Determine the fourth identification service, and then , the image recognition management service obtains the currently loaded atlas information corresponding to the third recognition service from the database service, and notifies the fourth recognition service to load the atlas corresponding to the currently loaded atlas information in the fourth recognition service, wherein the current The loaded atlas information can be any information related to identifying the atlas currently loaded by the service, for example, whether the atlas is currently loaded, the number of atlas currently loaded, the identification information of each atlas currently loaded, each atlas currently loaded Any information such as the identification images in the atlas and the number, the index file corresponding to the identification images in each currently loaded atlas, and the address of the index file. It should be noted here that the "loading atlas" in the context refers to one or more index files corresponding to one or more recognition maps in the recognition service loading atlas. Loading the atlas corresponding to the atlas information includes loading, in the fourth identification service, an index file corresponding to the identification graph in the atlas corresponding to the currently loaded atlas information.

In some embodiments, the determining a fourth identification service in the at least one identification service includes: if the image identification management service receives a message sent by a newly started fifth identification service in the at least one identification service registration information, and the fifth recognition service is determined as the fourth recognition service through the image recognition management service. In some embodiments, if the image recognition management service does not receive the heartbeat information sent by the third recognition service within a predetermined time interval, the image recognition management service waits for the new recognition service to start and registers with the image recognition management service, and registers the new start The fifth recognition service is used as the fourth recognition service, and then the image recognition management service reads the currently loaded atlas information corresponding to the third recognition service from the database service, and notifies the fourth recognition service to load the current image in the fourth recognition service. Load the atlas corresponding to the atlas information. For example, the registered recognition services in the image recognition management service include M1, M2, and M3. Each recognition service will send heartbeat information to the image recognition management service every 30 seconds. When the image recognition management service does not receive a message from M1 every 30 seconds When the heartbeat information is received, the image recognition management service determines that M1 is down; after that, if the image recognition management service receives the registration information sent by the newly started recognition service M4, the recognition service M4 is determined as the fourth recognition service, and then M4 loads M1 The atlas corresponding to the currently loaded atlas information is loaded, for example, the index file corresponding to the identification map in the atlas corresponding to the currently loaded atlas information is loaded.

In some embodiments, the determining of a fourth recognition service among the at least one recognition service includes: obtaining, through the image recognition management service, each activated one of the at least one recognition service from the database service The currently loaded atlas information of the recognition service; the image recognition management service determines a fourth recognition service according to the currently loaded atlas information of each activated recognition service, wherein the fourth recognition service satisfies at least the following One item: currently no atlases are loaded; the least number of currently loaded atlases; the least number of recognized images in the currently loaded atlases. In some embodiments, the currently loaded atlas information may be any information related to identifying the atlas currently loaded by the service, for example, whether the atlas is currently loaded, the number of currently loaded atlas, each currently loaded atlas , the identification information of each currently loaded atlas, and the number of identification images in each currently loaded atlas, the index file corresponding to the identification image in each currently loaded atlas, the address of the index file, and any other information. For example, the registered recognition services in the image recognition management service include M1, M2, and M3. Each recognition service will send heartbeat information to the image recognition management service every 30 seconds. When the image recognition management service does not receive a message from M1 every 30 seconds When the heartbeat information is received, the image recognition management service determines that M1 is down, then obtains the currently loaded atlas information corresponding to the currently started M2 and M3 from the database service, and then selects the currently loaded atlas from M2 and M3. M3 is used as the fourth identification service, or M3 with a smaller number of identification images in the loaded atlas is selected as the fourth identification service.

In some embodiments, the method further includes: for the image service request to be processed, if the target business service corresponding to the image service request to be processed has been deployed on multiple container sets, using the container management platform according to the Load information of each container set in multiple container sets, determine a target container set from the multiple container sets, and process the to-be-processed image service request through the target business service deployed on the target container set. In some embodiments, the image service request includes, but is not limited to, a publishing request for an image publishing service, a recognition request for an image recognition service, etc. For each image service request to be processed, a plurality of Different business services work together to complete this image service request, where each business service in multiple different business services can be deployed on a container set in the container cloud, or can be deployed in multiple different containers in the container cloud set. In some embodiments, for each image service request to be processed, multiple business services for processing this image service request are obtained first. For each business service in the multiple business services, if the business service is deployed in multiple For different container sets, it is necessary to select a target container set with a smaller load from multiple container sets according to the load information of each container set, where the load information can be the CPU load or memory load of the container set, and pass the target container set through the target container set. The business service deployed on the set is used to process this image service request, so as to achieve the purpose of load balancing. For example, the image service request to be processed is a publishing request of an image publishing service. In some embodiments, the request includes a publishing service, an indexing service, and an identification service. If the container set of the publishing service, the indexing service, and the identification service is deployed on the container cloud The numbers are two, three, and one, respectively, according to the load information of the two container sets where the publishing service is deployed, the target container set (target publishing service) with the smallest load information is determined. Similarly, according to the three container sets where the indexing service is deployed Load information of the container set, determine the target container set (target indexing service) with the smallest load information, and process this publishing request through the target publishing service, target indexing service and identification service.

In some embodiments, the method further includes: monitoring, through the container management platform, the running status information of the created one or more container sets, and if the monitored load information of the target container set meets a predetermined capacity expansion condition, creating one or more container sets. multiple second target container sets, and deploy business services corresponding to the target container sets on the one or more second target container sets. In some embodiments, the container management platform monitors the load information of each container set in one or more container sets that have been created on the container cloud, and if the monitored load information corresponding to a certain target container set satisfies a predetermined capacity expansion condition (For example, the load information corresponding to the target container set exceeds a predetermined load threshold), create one or more second target container sets, and deploy the business service corresponding to the target container set on the one or more second target container sets , thereby reducing the load on the target container set through automatic scaling. For example, if the monitored CPU load of the target container set exceeds 80%, a second target container set will be automatically expanded to deploy and run the business services corresponding to the target container set, so as to reduce the CPU load of the target container set and solve large-scale image problems. Identify technical issues with automatic scaling of services. If there are currently multiple identification requests for images to be identified, the feature extraction service is required to perform feature extraction on the multiple images to be identified. When the load information of the container set where the current feature extraction service is located meets the predetermined expansion conditions, the container management platform expands One or more container sets (each container set deploys a feature extraction service) and start one or more feature extraction services to process the feature extraction of multiple images to be identified, so as to achieve load balancing.

In some embodiments, the method further includes: in response to the expansion instruction for the target container set, creating one or more second target container sets through the container management platform, and creating one or more second target container sets in the one or more second target containers A business service corresponding to the target container set is deployed on the container set. In some embodiments, the expansion instruction includes, but is not limited to, for which target container set (corresponding to the business service) the expansion operation is performed and for which target container set is expanded several second target container sets. In some embodiments, in response to the expansion instruction for the target container set, one or more second target container sets are created on the container cloud, and the one or more second target container sets corresponding to the target container set are deployed on the one or more second target container sets business services, so as to reduce the load of the target container set through manual expansion, thereby solving the technical problem of rapid expansion of large-scale image recognition services.

In some embodiments, the method further includes: if the container management platform determines that the target user atlas satisfies a predetermined concurrent loading condition, determining one or more concurrent identification services corresponding to the target user atlas; Each of the one or more concurrent identification services loads the target user atlas respectively. In some embodiments, if the container management platform determines that the target user atlas satisfies a predetermined high concurrent loading condition (eg, the concurrent access volume of the target user atlas exceeds a predetermined concurrent access volume threshold), one or more concurrent identification services are allocated Loading the target user atlas at the same time can solve the problem of high concurrency of the specified atlas.

In some embodiments, the determining one or more concurrent identification services corresponding to the target user atlas includes: determining one corresponding to the target user atlas from identification services deployed in the at least one identification service or multiple concurrent recognition services. In some embodiments, one or more concurrent identification services for loading the target user atlas are determined from at least one identification service currently deployed on the container cloud. For example, the at least one identification service currently deployed on the container cloud may be The load status of each recognition service of , take one or more recognition services with smaller loads as concurrent recognition services, and load the target user atlas. For another example, according to the currently loaded atlas information of each identification service in the at least one identification service currently deployed on the container cloud, one or more identification services with a smaller number of index files corresponding to the loaded identification diagram are used as concurrent identification services. , and load the target user atlas.

In some embodiments, the determining one or more concurrent identification services corresponding to the target user atlas includes: creating one or more second target container sets, in the one or more second target container sets One or more sixth identification services are respectively deployed on each second target container set, and the one or more sixth identification services are used as one or more concurrent identification services corresponding to the target user atlas. In some embodiments, one or more second target container sets are created on the container cloud, and one or more sixth target container sets are respectively deployed on each of the one or more second target container sets. Identify the service, use the one or more sixth identification services as concurrent identification services, and load the target user atlas, where only one sixth identification service may be deployed on each second target container set, or, it may also be A plurality of sixth identification services are respectively deployed on each second target container set.

In some embodiments, the method further includes dividing the target user atlas into a plurality of mutually disjoint subsets if the container management platform determines that the target user atlas satisfies a predetermined division condition, wherein the multiple Each of the subsets includes at least one identification map; determining one or more identification services corresponding to the target user atlas; respectively loading the plurality of subsets through each of the one or more identification services A subset of , where different identification services correspond to different subsets. In some embodiments, if it is determined that the target user atlas satisfies a predetermined division condition (for example, the number of identification images in the target user atlas exceeds a predetermined number threshold), the target user atlas is divided into a plurality of mutually disjoint subsets, Each subset includes at least one identification map, and the container management platform determines that multiple identification services load a different subset respectively, so as to solve the problem that the number of atlas and pictures increases and the number of containers is insufficient. In some embodiments, the determining one or more identification services corresponding to the target user atlas includes: determining one or more identification services corresponding to the target user atlas from at least one deployed identification service ; In other embodiments, the determining one or more identification services corresponding to the target user atlas includes: creating one or more second target container sets, in the one or more second target containers One or more identification services are deployed on the set, and the one or more identification services are used as one or more identification services corresponding to the target user atlas.

Fig. 3 shows a flowchart of an image publishing method based on an image recognition service deployed on a container cloud according to an embodiment of the present application, wherein an image recognition service corresponding to the image recognition service is deployed on the container cloud, and the The container management platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: front-end resource service, indexing service, at least one identification service, content management service, and publishing service. For each business service in the image recognition business service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the business service. The mapping relationship between service names and service addresses of other dependent business services, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, the method includes steps S21, S22, and S22. S23 and step S24. In step S21, receiving an image publishing service request for a target recognition map through the publishing service, obtaining the target recognition map, and providing the target recognition map to the indexing service according to the service address of the indexing service; In step S22, the target identification map is obtained through the indexing service, a target index file corresponding to the target identification map is generated, and the target storage address corresponding to the target index file is returned to the publishing service; in step S23 , obtain the target storage address through the publishing service, and provide the target storage address to the target recognition service corresponding to the target atlas corresponding to the target recognition map in the at least one recognition service; in step S24 , obtain the target storage address through the target identification service, and load the target index file according to the target storage address.

In some embodiments, "publishing" in the context refers to adding an object recognition map to a recognition server or loading an index file corresponding to the object recognition map to the recognition server to provide retrieval. For example, the user uploads the target recognition map in the front-end resource service interface, and performs a series of operations on the target recognition map, such as adding the target recognition map to an existing atlas or creating a new atlas and placing the material images in the new atlas , another example is to bind the target recognition map to an AR material, and the content management service obtains the target recognition map and the atlas information corresponding to the target recognition map in response to the relevant operation information performed by the user on the target recognition map. In some embodiments , you can also obtain the AR material bound to the target recognition map, and update the relevant information of the target recognition map in the database service based on this, where the related information includes any of the following: the identification information of the target recognition map, The storage address of the target recognition map, the identification information of the atlas corresponding to the target recognition map, the identification information of the AR material bound to the target recognition map, the storage address of the AR material bound to the target recognition map, etc., to generate the target recognition map. The image publishing service request is sent to the publishing service. In some embodiments, the image publishing service request includes a target recognition map, wherein the target recognition map can be the target recognition map itself or the identification information of the target recognition map. , such as name, ID, etc. Alternatively, the image publishing service request may include, in addition to the target identification map, information of the atlas corresponding to the target identification map (for example, the identification of the atlas, etc.). In some embodiments, when the publishing service is started, a configuration file corresponding to the publishing service is loaded from the configuration service according to the configuration service address in the deployment file. The configuration file includes the service address of the indexing service on which the publishing service depends, and the publishing service receives After receiving the image publishing service request for the target recognition map, obtain the target recognition map corresponding to the image publishing service request (it can be the target recognition map itself, or the identification information of the target recognition map, such as name, ID, etc.), and according to the configuration file The service address of the indexing service in , provides the target identification map or the identification information of the target identification map to the indexing service, and the local indexing service or cloud indexing service generates the index file corresponding to the target identification map (the local indexing service is used to generate the index of local identification. The cloud index service is used to generate the index file identified by the cloud, and the feature extraction algorithms corresponding to the two index services can be the same or different), the generated index file will be stored in the file storage service, and the index file corresponding to the index file will be stored. The address is returned to the publishing service, and the publishing service provides the storage address to the identification service corresponding to the atlas corresponding to the target identification map, and the identification service loads the corresponding index file from the file storage service to the memory according to the storage address of the index file to provide the target A recognition service that recognizes graphs. In some embodiments, after the recognition service loads the index file, the relevant information corresponding to the target recognition map in the database service is updated, such as updating the publication status of the target recognition map, updating the recognition server information corresponding to the atlas corresponding to the target recognition map, etc. . Among them, the publishing service can obtain the atlas information corresponding to the target recognition map through the information contained in the image publishing service request, and can also obtain the information corresponding to the target recognition map through the publishing service to the database service. The information of the atlas is not limited here. In some embodiments, the business service in the image recognition business service further includes an account service, and the account service is used to provide services related to account management, such as managing users, enterprise-related information, and the like.

In some embodiments, the method further includes, before the step S21: receiving, through the front-end resource service, the user's operation information on the target identification map, and forwarding the operation information to a content management service; through the content management The service obtains the target recognition map and the target atlas set corresponding to the target recognition map according to the operation information, and updates the relevant information of the target recognition map in the database service in the basic service corresponding to the image recognition service, An image publishing service request for the object recognition map is generated and sent to the publishing service. In some embodiments, the user uploads the target recognition map in the front-end resource service interface, and performs a series of operations on the target recognition map, such as adding the target recognition map to an existing atlas or creating a new atlas and placing the target recognition map Enter the new atlas, or bind the target recognition map to an AR material, the front-end resource service receives the relevant operation information performed by the user on the target recognition map, and forwards the relevant operation information to the content management service, the content management service According to the received relevant operation information, the target recognition map and the atlas corresponding to the target recognition image are obtained, and in some embodiments, the AR material bound to the target recognition map can also be obtained, and the target is updated in the database service accordingly. Relevant information of the recognition map, wherein the related information includes any of the following: the identification information of the target recognition map, the storage address of the target recognition map, the identification information of the atlas corresponding to the target recognition map, the binding of the target recognition map The identification information of the AR material, the storage address of the AR material bound to the target recognition map, etc., generate an image publishing service request for the target recognition map and send it to the publishing service. In some embodiments, the image publishing service request may only include an object identification map, wherein the object identification map may be the object identification map itself, or may be identification information of the object identification map, such as name, ID, and the like. Alternatively, the image publishing service request may also include the target identification map and the information of the atlas corresponding to the target identification map (for example, the identification of the atlas, etc.).

In some embodiments, the indexing service obtains the target identification map, generates a target index file corresponding to the target identification map, and returns the target storage address corresponding to the target index file to the publishing service, including: The indexing service obtains the target identification map, generates a target index file corresponding to the target identification map, and provides the target index file to the file storage service in the basic service; obtains the target index file through the file storage service. the target index file, store and obtain the target storage address corresponding to the target index file, and return the target storage address to the indexing service; obtain the target storage address through the indexing service, and store the target The address returned to the publishing service. In some embodiments, the indexing service generates an index file corresponding to the target recognition picture after obtaining the target recognition map provided by the publishing service (which may be the target recognition map itself, or the identification information of the target recognition map, etc.) The index file is provided to the file storage service in the basic service for storage. After the index file is stored, the file storage service returns the corresponding storage address to the index service, and then the index service returns the storage address to the publishing service. In some embodiments, the business service in the image recognition business service further includes a resource storage service, the resource storage service is used to provide an intermediate service for file storage, encapsulates a standard storage interface to other services, and can be connected to multiple services later. A file storage service. The resource storage service is equivalent to an adapter, which is used to adapt to different file storage formats, such as different file storage services such as Qiniu Cloud, Xiaomi Cloud, and fastdfs. In some embodiments, the indexing service providing the target index file to the file storage service in the basic service further includes that the indexing service first converts the target index file by using the resource storage service, and then converts the converted index file The file is sent to the file storage service for storage.

In some embodiments, the providing the target storage address to the target recognition service corresponding to the target atlas set corresponding to the target recognition map in the at least one recognition service through the publishing service includes: The publication service obtains the publication status indication information corresponding to the target atlas from the database service in the basic service corresponding to the image recognition service; if the publication status indication information indicates that the target atlas has been published, the publication The service obtains the target recognition service corresponding to the target atlas in the at least one recognition service from the database service, and provides the target storage address to the target atlas corresponding to the target atlas according to the service address of the target recognition service. a target recognition service; if the publication status information indicates that the target atlas is not published, determine a target recognition service from the deployed recognition services in the at least one recognition service through the publication service, and determine the target recognition service according to the target recognition service provides the target storage address to the target identification service. In some embodiments, the publication status indication information is used to indicate the publication status of the target atlas, including but not limited to published, unpublished, published, etc. The publication service obtains the corresponding target atlas by querying the database service. Publishing status, if the target atlas has been published, query the database service to obtain the target identification service corresponding to the target atlas (it can be the identification information of the target identification service, such as name, address, etc.), and provide the storage address of the target index file. Provide the target identification service; if the target atlas is not published, determine the target identification service from at least one identification service, and provide the storage address of the target index file to the determined target identification service. In some embodiments, for the newly added unpublished atlas, the identification service with a small number of loaded pictures/low load is given priority to publish, and for the published atlas, the information about the identification service is queried from the database service, and a notification is made. The corresponding recognition service loads the index file corresponding to the target recognition map.

In some embodiments, the determining, by the publishing service, the target recognition service from the deployed recognition services in the at least one recognition service comprises: obtaining the at least one recognition from the database service by the publishing service The currently loaded atlas information corresponding to the deployed identification services in the service; the target identification service is determined from the deployed identification services according to the currently loaded atlas information through the publishing service, wherein the target identification service The corresponding current loading information satisfies at least one of the following: the number of currently loaded atlases is the least; the number of recognition maps in the currently loaded atlas is the least. In some embodiments, the publishing service obtains the currently loaded atlas information for each of the at least one identification service from the database service, and then, according to the currently loaded atlas information for each identification service, obtains information from the at least one identification service. A target recognition service is determined in the service, wherein the target recognition service may be the recognition service with the least number of currently loaded atlases, or may be the recognition service with the least number of recognition graphs in the currently loaded atlases. In some embodiments, the currently loaded atlas information may be any information related to the atlas currently loaded by the target recognition service, such as whether the atlas is currently loaded, the number of currently loaded atlas, each currently loaded atlas , the identification information of each currently loaded atlas, and the number of identification images in each currently loaded atlas, the index file corresponding to the identification image in each currently loaded atlas, the address of the index file, and any other information. For example, if the recognition service includes M1, M2 and M3, the number of loaded atlases in M1 is 20, the number of loaded atlases in M2 is 32, and the number of loaded atlases in M3 is 25, then select the currently loaded atlas M1 with the smallest number is used as the target recognition service corresponding to the target atlas. For another example, the recognition service includes N1, N2 and N3, the number of recognition maps loaded in N1 is 450, the number of recognition maps loaded in N2 is 312, and the number of recognition maps loaded in N3 is 256, then select the currently loaded recognition map N3 with the least number of graphs is used as the target recognition service corresponding to the target atlas. It should be noted that the above-mentioned conditions met by the target recognition service are only examples, not limitations of the present application. Any implementation method for determining the target recognition service corresponding to the target atlas should be included in the protection scope of the present application. In some embodiments, the business service in the image recognition business service further includes an image recognition management service, the image recognition management service is used to manage the at least one recognition service; wherein the method further comprises: for the Each recognition service in the at least one recognition service, when the recognition service is started, sends registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service; The management service receives the registration information, generates relevant information of the recognition service according to the registration information, and stores the relevant information of the recognition service in the database service in the basic service corresponding to the image recognition service. In some embodiments, the business service in the image recognition business service further includes an image recognition management service, the image recognition management service is used to generate the relevant information of the recognition service in the database service (for example, the currently loaded recognition map ID, index file address, atlas, publishing time and other currently loaded atlas information) to record the relevant information of each currently activated identification service, and the publishing service can obtain at least one identification service from the database service. The atlas information currently loaded by each identification service is identified, and then the target identification service is determined from the at least one identification service according to the currently loaded atlas information of each identification service.

In some embodiments, the method further includes: for each activated recognition service in the at least one recognition service, periodically sending heartbeat information to the image recognition management service through the activated recognition service; The image recognition management service does not receive the heartbeat information sent by the seventh recognition service in the at least one recognition service within the predetermined time interval, then the eighth recognition service is determined in the at least one recognition service, and the database service is obtained from the database service. Obtain currently loaded atlas information corresponding to the seventh identification service, so as to load the atlas corresponding to the currently loaded atlas information in the eighth identification service. The relevant operations in this embodiment have been described in detail in the foregoing embodiments, and are not repeated here.

In some embodiments, the method further includes: for the image publishing service request, if the target business service corresponding to the image publishing service request has been deployed on multiple container sets, using the container management platform according to the Load information of each container set in multiple container sets, determine a target container set from the multiple container sets, and process the image publishing service request through the target business service deployed on the target container set.

4 shows a flowchart of an image recognition method based on an image recognition service deployed on a container cloud according to an embodiment of the present application, wherein an image recognition service corresponding to the image recognition service is deployed on the container cloud, and the The container management platform manages the image recognition business service, and the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, image recognition API service, for the image recognition business service For each business service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains other business services that the business service depends on. The mapping relationship between the service name and the service address, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, the method includes steps S31, S32, S33, S34 and Step S35. In step S31, an image recognition service request for the picture to be recognized sent by the target application is received through the image recognition API service, and a user atlas corresponding to the to-be-recognized picture is obtained; in step S32, the image recognition The API service obtains the target recognition service corresponding to the user atlas in the at least one recognition service from the database service in the basic service corresponding to the image recognition service, and extracts the target recognition service according to the service address of the feature extraction service The to-be-recognized picture and the target recognition service are provided to the feature extraction service; in step S33, the to-be-recognized picture and the target recognition service are obtained through the feature extraction service, and feature information of the to-be-recognized picture is extracted , and provide the feature information to the target recognition service according to the service address of the target recognition service; in step S34, obtain the feature information through the target recognition service, and combine the feature information with the target The index file currently loaded by the identification service is matched; in step S35, if the matching is successful, the target identification map information corresponding to the target index file matching the feature information is obtained through the target identification service, and the target The recognition map information is returned to the image recognition API service, the target recognition map information is obtained through the image recognition API service, the target AR material corresponding to the target recognition map information is obtained from the basic service, and the target AR The material is returned to the target application.

In some embodiments, the target application sends the to-be-recognized image, and after the recognition API service receives the to-be-recognized image, a user atlas corresponding to the to-be-recognized image is obtained. For example, the atlas corresponding to the target application may be used as the to-be-recognized image The corresponding user atlas, or, the atlas corresponding to the user account can be used as the user atlas corresponding to the to-be-recognized picture, or the specified atlas information sent by the target application can also be used as the to-be-recognized The corresponding user atlas . In some embodiments, when the image recognition API service is started, a configuration file corresponding to the image recognition API service is loaded from the configuration service according to the configuration service address in the deployment file, and the configuration file includes the feature extraction service on which the image recognition API service depends. service address. In some embodiments, the image recognition API service queries the database service for the target recognition service corresponding to the user atlas (eg, the address, name, etc. of the target recognition service). Further, the business service in the image recognition business service also includes a publishing service, the image recognition API service calls the publishing service, and the publishing service queries the target recognition service (such as the target recognition service) corresponding to the user atlas from the database service. address, name, etc.), and returns the object recognition service to the image recognition API service. In other embodiments, the business service in the image recognition business service further includes an image recognition management service, and the image recognition management service is used to generate the relevant information of the recognition service (for example, the currently loaded recognition map) in the database service ID, index file address, atlas, release time and other currently loaded atlas information) to record the relevant information of each currently activated recognition service, and the image recognition API service can obtain the corresponding atlas from the database service. object recognition service, and then returns the object recognition service to the image recognition API service. The image recognition API service sends the image to be recognized and the corresponding target recognition service (such as the address and name of the target recognition service) to the feature extraction service according to the service address of the feature extraction service in the configuration file, and the feature extraction service extracts the to-be-recognized image. After identifying the feature information corresponding to the picture, according to the service address of the target recognition service, the feature information is provided to the target recognition service, and the target recognition service matches the feature information with one or more index files currently loaded by the target recognition service, If the matching is successful, obtain the target identification map information (for example, identification map ID, identification map name, etc.) corresponding to the target index file matching the feature information, and return the target identification map information to the image recognition API service, image recognition The API service queries the basic service according to the returned target recognition map information to obtain the AR material bound to the target recognition map information, and the image recognition API service returns the AR material to the target application. The image recognition API service returns the AR material to the target application. In some embodiments, the image recognition API service may obtain the AR material from the file storage service in the basic service and directly convert the AR material to the target application. Return to the target application, or, the image recognition API service can also obtain the storage address corresponding to the AR material from the database service and return the download address corresponding to the AR material to the target application, and the target application downloads the corresponding AR material, here Not limited; in some embodiments, if the matching fails, the target recognition service or the image recognition API service generates image recognition service response information for indicating that the image to be recognized fails to be recognized, and returns the image recognition service response information to the target application, to prompt the user that the identification failed. In some embodiments, the business service in the image recognition business service further includes an account service, and the account service is used to provide services related to account management, such as managing users, enterprise-related information, and the like.

In some embodiments, the receiving, by using the image recognition API service, an image recognition service request for a picture to be recognized and sent by a target application, and obtaining a user atlas corresponding to the to-be-recognized picture, includes: using the image recognition The API service receives the image recognition service request sent by the target application for the picture to be recognized, and performs access authority authentication on the target application; if the authentication is passed, the user atlas corresponding to the to-be-recognized picture is obtained. In some embodiments, the target application sends the to-be-recognized picture to the recognition API service, and the image recognition API service will first perform access authorization authentication on the target application or user identity. The corresponding user atlas, and identify the image to be identified. In some embodiments, the method further includes: for the image recognition service request, if the target business service corresponding to the image recognition service request has been deployed on multiple container sets, using the container management platform according to the Load information of each container set in multiple container sets, determine a target container set from the multiple container sets, and process the image recognition service request through the target business service deployed on the target container set.

In some embodiments, the business service in the image recognition business service further includes at least one of the following: a front-end resource service, an indexing service, a content management service, and a resource storage service, wherein the front-end resource service is used to receive user requests for Relevant operation information performed by the target recognition map; the content management service responds to the relevant operation information performed by the user on the target recognition map, and accordingly updates the relevant information of the target recognition map in the database service; the index service user generates the target Identify the index file of the graph; the resource storage service is used to provide an intermediate service for file storage, encapsulates a standard storage interface to other services, and can be connected to a variety of file storage services later.

In some embodiments, the business service in the image recognition business service further includes an image recognition management service, the image recognition management service is used to manage the at least one recognition service; wherein the method further comprises: for the Each recognition service in the at least one recognition service, when the recognition service is started, sends registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service; The management service receives the registration information, generates relevant information of the recognition service according to the registration information, and stores the relevant information of the recognition service in the database service in the basic service corresponding to the image recognition service. The relevant operations in this embodiment have been described in detail in the foregoing embodiments, and are not repeated here.

In some embodiments, the method further includes: for each activated recognition service in the at least one recognition service, periodically sending heartbeat information to the image recognition management service through the activated recognition service; The image recognition management service does not receive the heartbeat information sent by the ninth recognition service in the at least one recognition service within a predetermined time interval, then the tenth recognition service is determined in the at least one recognition service, and the database service is obtained from the database service. Obtain currently loaded atlas information corresponding to the ninth identification service, so as to load the atlas corresponding to the currently loaded atlas information in the tenth identification service. The relevant operations in this embodiment have been described in detail in the foregoing embodiments, and are not repeated here.

FIG. 5 shows a structural diagram of a network device for deploying an image recognition service on a container cloud according to an embodiment of the present application. The device includes a first module 11 and a second module 12 . A module 11 is configured to obtain a deployment file corresponding to the image recognition business service in the image recognition service after the basic service in the image recognition service is deployed, wherein the deployment file includes the image recognition business service in the The service name, image file address, number of container sets, and configuration service address corresponding to each business service. The business services in the image recognition business service include at least one of the following: image recognition management service, front-end resource service, indexing service, at least one A recognition service, feature extraction service, account service, content management service, publishing service, image recognition API service, and resource storage service; one or two modules 12 are used for each business service in the image recognition business service, according to the The deployment file creates a container set on the container cloud and starts the business service, loads the corresponding configuration file according to the configuration service address corresponding to the business service through the container set, and manages the business service through the container management platform. The configuration file contains the mapping relationship between the service names and service addresses of other business services that the business service depends on, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, by The image recognition business service deployed by the container cloud performs at least any one of the following: receiving an image publishing service request for a first recognition map through the publishing service, obtaining the first recognition map, and publishing the first recognition map. The identification map is provided to the indexing service; the first identification map is obtained through the indexing service, a first index file corresponding to the first identification map is generated, and the first storage address corresponding to the first index file is stored Return to the publishing service; obtain the first storage address through the publishing service, and provide the first storage address to the first atlas corresponding to the first identification map in the at least one identification service obtain the first storage address through the first identification service, and load the first index file according to the first storage address; or, through the image recognition API service to receive the data sent by the target application , for the image recognition service request of the picture to be recognized, obtain the user atlas corresponding to the picture to be recognized; obtain the at least one recognition service from the database service in the basic service through the image recognition API service. The second identification service corresponding to the user atlas, provides the picture to be identified and the second identification service to the feature extraction service; obtains the picture to be identified and the second identification service through the feature extraction service , extract the feature information of the picture to be recognized, and provide the feature information to the second recognition service; obtain the feature information through the second recognition service, and combine the feature information with the second recognition service The index file currently loaded by the service is matched; if the matching is successful, the second identification map information corresponding to the second index file matching the feature information is obtained through the second identification service, and the second identification map The information is returned to the image recognition API service, through the image recognition A The PI service obtains the second identification map information, obtains the second AR material corresponding to the second identification map information from the basic service, and returns the second AR material to the target application.

In some embodiments, after the basic service in the image recognition service is deployed, k8s obtains a deployment file corresponding to the image recognition business service in the image recognition service. The deployment file is the basic file for starting the image recognition service, and k8s is specified in it. The required basic content, such as specifying the service name of each business service in the image recognition business service, the image file address, the number of container sets, and the environment variables when the container set is running (such as the address of the configuration service, etc.) in the deployment file, Further, the configuration item key can also be specified in the environment variable, and further, the content of the service (SVC) of each business service (including the exposed port and other content) can also be specified in the deployment file, and k8s is created according to the deployment file. Container set and start each business service in the image recognition business service on the container set. The image recognition business service includes at least one of the following: image recognition management service, front-end resource service, indexing service, recognition service, feature extraction service, account service , content management service, publishing service, image recognition API service, resource storage service, wherein, there may be one or more deployment files, which are not limited here. In some embodiments, each business service corresponds to the same configuration service address, and in other embodiments, each business service corresponds to a different configuration service address, which is not limited herein. There is no restriction on the deployment sequence of different business services here. In some embodiments, the image recognition management service is used to perform service control, such as controlling the allocation of recognition servers, and the like. The front-end resource service is used to manage the back-end or the front-end of the web page, and is also an entry service included in the corresponding client. In some embodiments, divided according to business logic, the image recognition business service includes a second layer service that depends on the basic service and a first layer service that depends on the second layer service, wherein the second layer service The service is used to provide services that are biased towards the underlying algorithm function type (such as at least one service such as indexing service, identification service, feature extraction service, etc.), and has little relevance to specific business logic. The first layer service is used to provide Services that are biased towards upper-level business logic types (such as account services, content management services, publishing services, image recognition API services, etc.) are closely related to specific business logic. Among them, it should be noted that the first First-tier services and second-tier services are divided according to business logic, which does not mean that there is hierarchical division in deployment, but is only used to facilitate understanding of business logic. Dependency means that the relying party directly or indirectly uses or invokes or updates the data, functions, interfaces, etc. of the relying party. For example, the publishing service in the first-tier service needs to call the indexing service in the second-tier service; for example, the first-tier service The content management service in the service updates the database content in the base service, etc. Figure 2 is a service architecture diagram of an example of the application, in which all business services and configuration editors are stored in the mirror warehouse, the mirror address is configured after the service is created in the container cloud, and the corresponding configuration is obtained from the configuration service (etcd) after the service is started. , initialize the mysql database tables and data, and complete the construction of the service architecture; among them, the cluster service entrance is the front-end resource service (webhiar), and the service request is forwarded to the account service (accountsrv), content management service (contentsrv), publishing service (publishsrv), The first layer of services such as image recognition API service (recognizesrv), the first layer relies on indexing services (including cloud indexing service (cloudkey), local indexing service (localkey)), recognition service (searchsrv), feature extraction service (featuresrv) and other third-level services. The second-tier service, the second-tier service relies on basic services such as the database service mysql, the file storage service fastdfs, and the resource storage service (storagesrv). Among them, webhiar is used as the access entry, which is used to deploy the entry service of the API of the management background or the front end of the web page. The access of the application and the API request of the application end are all entrances from this node. When concurrent access occurs, the server can be deployed on the server. After building load balancing before, different services in the architecture can expand nodes according to the pressure of different requests. It should be noted that, the various services involved in the above will be described in detail in subsequent embodiments. In some embodiments, by deploying the image recognition service on a container cloud, the portability and security of the image recognition service can be improved.

In some embodiments, for each business service in the image recognition business service, a container is created on the container cloud according to deployment information such as the service name, image file address, number of container sets, etc. corresponding to each business service in the deployment file set and start the business service, the container set loads the configuration file corresponding to the business service from the configuration service according to the configuration service address in the deployment file (for example, the configuration service includes a plurality of configuration files corresponding to different business services respectively , the configuration file corresponding to the business service can be found in the configuration service through the service name of the business service and loaded) to complete the startup of the business service, and the container management platform (for example, k8s) has been deployed on the container cloud. At least one of the image recognition business services is managed. In some embodiments, a configuration file needs to be loaded when the business service is started, and the configuration file contains the content required by the business service when it runs, usually including the port number occupied by the business service, the database address of the connection, the address of the cache server, dependencies (need to call ) service address of other business services. In some embodiments, K8S creates a container set corresponding to the business service and the number of container sets on the container cloud according to the number of container sets corresponding to the business service in the deployment file, and according to the image file corresponding to the business service in the deployment file address, download the image file corresponding to the business service, start the business service on the container set according to the image file, and then set the environment variable according to the deployment file, and the container set reads the configuration service set in the environment variable. address, load the configuration file corresponding to the business service from the configuration service, and complete the startup of the business service according to the content of the configuration file, and then the business service can process the business request.

In some embodiments, if the basic service is deployed on the container cloud, at least one of the basic service and the image recognition business service deployed on the container cloud can be managed through the container management platform. In other embodiments, if the basic service is deployed on the peripheral service of the container cloud, at least one service among the image recognition business services deployed on the container cloud can be managed through the container management platform. In some embodiments, the k8s master node (Master Node) accepts instructions from the outside, and then forwards the instructions to the attached Node (node), and then allocates resources on the node and assigns Pod (container set) to complete the task request. Each service is encapsulated in a docker image and runs in each Pod in k8s. In some embodiments, using k8s to manage the image recognition service deployed on the container cloud can solve the technical problem of rapid expansion of large-scale image recognition services, realize load balancing and dynamic expansion of image recognition services, reduce deployment difficulty, and increase Deployment flexibility to accelerate the launch of related products.

1) Indexing service

2) Identify the service

3) Feature extraction service

Here, the related layer 2 services are the same as or similar to the embodiment shown in FIG. 1 , so they are not described again, and are incorporated herein by reference.

1) Content Management Services

2) Publish service

3) Image recognition API service

4) Account service

Here, the related first-layer services are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, and are incorporated herein by reference.

In some embodiments, the basic service includes at least one of the following:

1) Configure the service

2) Database service

3) File storage service

4) Cache database service

5) Mirror warehouse service

6) Configure the editor service

Here, the related basic services are the same as or similar to the embodiment shown in FIG. 1 , so they will not be described again, but are incorporated herein by reference.

In some embodiments, the device is further configured to: deploy a resource storage service (storagesrv) on the container cloud. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, for each business service in the image recognition business service, creating a container set on the container cloud according to the deployment file and starting the business service includes: for each business service in the image recognition business service a business service, according to the number of container sets corresponding to the business service in the deployment file, create a container set corresponding to the business service and the number of container sets on the container cloud, and according to the number of container sets corresponding to the business service in the deployment file Image file address, download the image file corresponding to the business service, and start the business service on the container set according to the image file. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the image recognition management service is used to manage the at least one recognition service; wherein the device is further configured to: for each recognition service of the at least one recognition service, start the recognition service at the When the image recognition management service is used, send registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service; receive the registration information through the image recognition management service, and generate the identification according to the registration information The relevant information of the service is stored, and the relevant information of the identification service is stored in the database service in the basic service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: for each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service; if The image recognition management service does not receive the heartbeat information sent by the third recognition service in the at least one recognition service within the predetermined time interval, then determines the fourth recognition service in the at least one recognition service, The database service in the service obtains the currently loaded atlas information corresponding to the third identification service, so as to load the atlas corresponding to the currently loaded atlas information in the fourth identification service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the determining a fourth identification service in the at least one identification service includes: if the image identification management service receives a message sent by a newly started fifth identification service in the at least one identification service registration information, and the fifth recognition service is determined as the fourth recognition service through the image recognition management service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the determining of a fourth recognition service among the at least one recognition service includes: obtaining, through the image recognition management service, each activated one of the at least one recognition service from the database service The currently loaded atlas information of the recognition service; the image recognition management service determines a fourth recognition service according to the currently loaded atlas information of each activated recognition service, wherein the fourth recognition service satisfies at least the following One item: currently no atlases are loaded; the least number of currently loaded atlases; the least number of recognized images in the currently loaded atlases. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: for the image service request to be processed, if the target business service corresponding to the image service request to be processed has been deployed on multiple container sets, use the container management platform according to the Load information of each container set in multiple container sets, determine a target container set from the multiple container sets, and process the to-be-processed image service request through the target business service deployed on the target container set. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: monitor the running status information of the created one or more container sets through the container management platform, and if the monitored load information of the target container set meets the predetermined capacity expansion condition, create a or multiple second target container sets, and deploy business services corresponding to the target container sets on the one or more second target container sets. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: in response to the expansion instruction for the target container set, create one or more second target container sets through the container management platform, and create one or more second target container sets in the one or more second container sets A business service corresponding to the target container set is deployed on the target container set. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: if the container management platform determines that the target user atlas satisfies a predetermined concurrent loading condition, determine one or more concurrent identification services corresponding to the target user atlas; Each of the one or more concurrent identification services separately loads the target user graph. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the determining one or more concurrent identification services corresponding to the target user atlas includes: determining one corresponding to the target user atlas from identification services deployed in the at least one identification service or multiple concurrent recognition services. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the determining one or more concurrent identification services corresponding to the target user atlas includes: creating one or more second target container sets, in the one or more second target container sets One or more sixth identification services are respectively deployed on each second target container set, and the one or more sixth identification services are used as one or more concurrent identification services corresponding to the target user atlas. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

In some embodiments, the device is further configured to: if the container management platform determines that the target user atlas satisfies a predetermined division condition, divide the target user atlas into a plurality of mutually disjoint subsets, wherein the Each subset in the plurality of subsets includes at least one identification map; one or more identification services corresponding to the target user atlas are determined; the multiple identification services are respectively loaded through each of the one or more identification services A subset of the subsets, wherein different identification services correspond to different subsets. Here, the related operations are the same as or similar to the embodiment shown in FIG. 1 , so they are not repeated here, but are incorporated herein by reference.

FIG. 6 shows a network device structure diagram for image publishing based on an image recognition service deployed on a container cloud according to an embodiment of the present application, wherein an image recognition service corresponding to the image recognition service is deployed on the container cloud, and The image recognition business service is managed through the container management platform, and the business services in the image recognition business service include at least one of the following: front-end resource service, indexing service, at least one identification service, content management service, and publishing service. For each business service in the image recognition business service, create a container set on the container cloud according to the deployment file and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the business service. The mapping relationship between the service name and the service address of the other business services it depends on, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, the device includes two modules 21, two The second module 22 , the second third module 23 and the second fourth module 24 . The second module 21 is configured to receive an image publishing service request for a target recognition map through the publishing service, obtain the target recognition map, and provide the target recognition map to the index according to the service address of the index service The second module 22 is configured to obtain the target identification map through the indexing service, generate a target index file corresponding to the target identification map, and return the target storage address corresponding to the target index file to the publishing service The second and third modules 23 are used to obtain the target storage address through the publishing service, and provide the target storage address to the target corresponding to the target atlas corresponding to the target identification map in the at least one identification service. identification service; the second and fourth module 24 is configured to obtain the target storage address through the target identification service, and load the target index file according to the target storage address.

In some embodiments, "publishing" in the context refers to adding an object recognition map to a recognition server or loading an index file corresponding to the object recognition map to the recognition server to provide retrieval. For example, the user uploads the target recognition map in the front-end resource service interface, and performs a series of operations on the target recognition map, such as adding the target recognition map to an existing atlas or creating a new atlas and placing the target recognition map in the new map Another example is to bind the target recognition map to an AR material, and the content management service obtains the target recognition map and the atlas information corresponding to the target recognition map in response to the relevant operation information performed by the user on the target recognition map. In some implementations In an example, the AR material bound to the target recognition map can also be obtained, and the related information of the target recognition map can be updated in the database service based on this, wherein the related information includes any one of the following: identification information of the target recognition map , the storage address of the target recognition map, the identification information of the atlas corresponding to the target recognition map, the identification information of the AR material bound to the target recognition map, the storage address of the AR material bound to the target recognition map, etc. The image publishing service request for the image is sent to the publishing service. In some embodiments, the image publishing service request includes a target recognition map, where the target recognition map can be the target recognition map itself or the identification of the target recognition map. Information, such as name, ID, etc., or the image publishing service request may include, in addition to the target identification map, information of the atlas corresponding to the target identification map (for example, the identification of the atlas, etc.). In some embodiments, when the publishing service is started, a configuration file corresponding to the publishing service is loaded from the configuration service according to the configuration service address in the configuration file. The configuration file includes the service address of the indexing service on which the publishing service depends, and the publishing service receives After receiving the image publishing service request for the target recognition map, obtain the target recognition map corresponding to the image publishing service request (it can be the target recognition map itself, or the identification information of the target recognition map, such as name, ID, etc.), and according to the configuration file The service address of the indexing service in , provides the identification information of the target identification map or material image to the indexing service, and the local indexing service or cloud indexing service generates the index file corresponding to the target identification map (the local indexing service is used to generate the locally identified index file. , the cloud index service is used to generate the index file identified by the cloud, and the feature extraction algorithms corresponding to the two index services can be the same or different), the generated index file will be stored in the file storage service, and the index file corresponding to the index file will be stored at the address. Return to the publishing service, the publishing service provides the storage address to the identification service corresponding to the atlas corresponding to the target identification map, and the identification service loads the corresponding index file from the file storage service to the memory according to the storage address of the index file to provide target identification. Graph recognition service. In some embodiments, after the recognition service loads the index file, the relevant information corresponding to the target recognition map in the database service is updated, such as updating the publication status of the target recognition map, updating the recognition server information corresponding to the atlas corresponding to the target recognition map, etc. . Among them, the publishing service can obtain the atlas information corresponding to the target recognition map through the information contained in the image publishing service request, and can also obtain the map corresponding to the target recognition map through the publishing service to the database service. The information of the set is not limited here. In some embodiments, the business service in the image recognition business service further includes an account service, and the account service is used to provide services related to account management, such as managing users, enterprise-related information, and the like.

In some embodiments, the device is further configured to: receive, through the front-end resource service, user operation information on the target identification map, and forward the operation information to a content management service; and through the content management service according to the operation information, obtain the target recognition map and the target atlas corresponding to the target recognition map, update the relevant information of the target recognition map in the database service in the basic service corresponding to the image recognition service, and generate the target recognition map for the target An image publishing service request for a graph is identified and sent to the publishing service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, the indexing service obtains the target identification map, generates a target index file corresponding to the target identification map, and returns the target storage address corresponding to the target index file to the publishing service, including: The indexing service obtains the target identification map, generates a target index file corresponding to the target identification map, and provides the target index file to the file storage service in the basic service; obtains the target index file through the file storage service. the target index file, store and obtain the target storage address corresponding to the target index file, and return the target storage address to the indexing service; obtain the target storage address through the indexing service, and store the target The address returned to the publishing service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, providing the target storage address to a target recognition service corresponding to the target atlas set corresponding to the target recognition map in the at least one recognition service by the publishing service includes: by the publishing The service obtains the publication status indication information corresponding to the target atlas from the database service in the basic service corresponding to the image recognition service; if the publication status indication information indicates that the target atlas has been published, the target atlas is published through the publication service. The database service obtains the target recognition service corresponding to the target atlas in the at least one recognition service, and provides the target storage address to the target recognition corresponding to the target atlas according to the service address of the target recognition service service; if the publication status information indicates that the target atlas has not been published, determine the target recognition service from the deployed recognition services in the at least one recognition service through the publication service, and identify the service of the service according to the target The address provides the target storage address to the target identification service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, the determining, by the publishing service, the target recognition service from the deployed recognition services in the at least one recognition service comprises: obtaining the at least one recognition from the database service by the publishing service The currently loaded atlas information corresponding to the deployed identification services in the service; the target identification service is determined from the deployed identification services according to the currently loaded atlas information through the publishing service, wherein the target identification service The corresponding current loading information satisfies at least one of the following: the number of currently loaded atlases is the least; the number of recognition maps in the currently loaded atlas is the least. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, the business service in the image recognition business service further includes an image recognition management service, and the image recognition management service is used to manage the at least one recognition service; wherein, the device is further configured to: for all Each recognition service in the at least one recognition service, when the recognition service is started, according to the domain name information and/or address information of the image recognition management service, send registration information to the image recognition management service; The recognition management service receives the registration information, generates relevant information of the recognition service according to the registration information, and stores the relevant information of the recognition service in the database service in the basic service corresponding to the image recognition service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, the device is further configured to: for each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service; if The image recognition management service does not receive the heartbeat information sent by the seventh recognition service in the at least one recognition service within a predetermined time interval, then determines the eighth recognition service in the at least one recognition service, and obtains the information from the database. The service obtains the currently loaded atlas information corresponding to the seventh identification service, so as to load the atlas corresponding to the currently loaded atlas information in the eighth identification service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

In some embodiments, the device is further configured to: for the image publishing service request, if the target business service corresponding to the image publishing service request has been deployed on multiple container sets, use the container management platform to load information of each container set in the multiple container sets, determine a target container set from the multiple container sets, and process the image publishing service request through the target business service deployed on the target container set. Here, the related operations are the same as or similar to the embodiment shown in FIG. 3 , so they are not described again, but are incorporated herein by reference.

FIG. 7 shows a network device structure diagram of image recognition based on an image recognition service deployed on a container cloud according to an embodiment of the present application, wherein an image recognition service corresponding to the image recognition service is deployed on the container cloud, and The image recognition business service is managed through the container management platform, and the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, and image recognition API service, for the image recognition service For each business service in the service, create a container set on the container cloud according to the deployment file and start the business service. The container set loads the configuration file corresponding to the business service, and the configuration file contains other services that the business service depends on. The mapping relationship between the service name of the service and the service address, so that the business service calls other business services that the business service depends on according to the mapping relationship; wherein, the device includes a three-one module 31, three-two modules 32, three Three modules 33 , three four modules 34 and three five modules 35 . The 31 module 31 is used to receive, through the image recognition API service, the image recognition service request for the picture to be recognized sent by the target application, and obtain the user atlas corresponding to the picture to be recognized; the 32 module 32 is used to pass The image recognition API service obtains the target recognition service corresponding to the user atlas in the at least one recognition service from the database service in the basic service corresponding to the image recognition service, and extracts the service of the service according to the feature address to provide the to-be-recognized picture and the target recognition service to the feature extraction service; the third module 33 is configured to obtain the to-be-recognized picture and the target recognition service through the feature extraction service, and extract the The feature information of the picture to be recognized, and the feature information is provided to the target recognition service according to the service address of the target recognition service; the third and fourth modules 34 are used to obtain the feature information through the target recognition service, and The feature information is matched with the index file currently loaded by the target recognition service; the three-five module 35 is used to obtain, through the target recognition service, the corresponding target index file that matches the feature information if the matching is successful. target recognition map information, and return the target recognition map information to the image recognition API service, obtain the target recognition map information through the image recognition API service, and obtain the target recognition map information from the basic service. target AR material, and return the target AR material to the target application.

In some embodiments, the target application sends the to-be-recognized image, and after the recognition API service receives the to-be-recognized image, a user atlas corresponding to the to-be-recognized image is obtained. For example, the atlas corresponding to the target application may be used as the to-be-recognized image The corresponding user atlas, or, the atlas corresponding to the user account can be used as the user atlas corresponding to the to-be-recognized picture, or the specified atlas information sent by the target application can also be used as the to-be-recognized The corresponding user atlas . In some embodiments, when the image recognition API service is started, a configuration file corresponding to the image recognition API service is loaded from the configuration service according to the configuration service address in the configuration file, and the configuration file includes the feature extraction service on which the image recognition API service depends. service address. In some embodiments, the image recognition API service queries the database service for the target recognition service corresponding to the user atlas (eg, the address, name, etc. of the target recognition service). Further, the business service in the image recognition business service also includes a publishing service, the image recognition API service calls the publishing service, and the publishing service queries the target recognition service (such as the target recognition service) corresponding to the user atlas from the database service. address, name, etc.), and returns the object recognition service to the image recognition API service. In other embodiments, the business service in the image recognition business service further includes an image recognition management service, and the image recognition management service is used to generate the relevant information of the recognition service (for example, the currently loaded recognition map) in the database service ID, index file address, atlas, release time and other currently loaded atlas information) to record the relevant information of each currently activated recognition service, and the image recognition API service can obtain the corresponding atlas from the database service. object recognition service, and then returns the object recognition service to the image recognition API service. The image recognition API service sends the image to be recognized and the corresponding target recognition service (such as the address and name of the target recognition service) to the feature extraction service according to the service address of the feature extraction service in the configuration file, and the feature extraction service extracts the to-be-recognized image. After identifying the feature information corresponding to the picture, according to the service address of the target recognition service, the feature information is provided to the target recognition service, and the target recognition service matches the feature information with one or more index files currently loaded by the target recognition service, If the matching is successful, obtain the target identification map information (for example, identification map ID, identification map name, etc.) corresponding to the target index file matching the feature information, and return the target identification map information to the image recognition API service, image recognition The API service queries the basic service according to the returned target recognition map information to obtain the AR material bound to the target recognition map information, and the image recognition API service returns the AR material to the target application. The image recognition API service returns the AR material to the target application. In some embodiments, the image recognition API service can obtain the AR material from the file storage service in the basic service and directly convert the AR material to the target application. Return to the target application, or, the image recognition API service can also obtain the storage address corresponding to the AR material from the database service and return the download address corresponding to the AR material to the target application, and the target application downloads the corresponding AR material, here Not limited; in some embodiments, if the matching fails, the target recognition service or the image recognition API service generates image recognition service response information for indicating that the image to be recognized fails to be recognized, and returns the image recognition service response information to the target application, to prompt the user that the identification failed. In some embodiments, the business service in the image recognition business service further includes an account service, and the account service is used to provide services related to account management, such as managing users, enterprise-related information, and the like.

In some embodiments, the receiving, by using the image recognition API service, an image recognition service request for a picture to be recognized and sent by a target application, and obtaining a user atlas corresponding to the to-be-recognized picture, includes: using the image recognition The API service receives the image recognition service request sent by the target application for the picture to be recognized, and performs access authority authentication on the target application; if the authentication is passed, the user atlas corresponding to the to-be-recognized picture is obtained. Here, the related operations are the same as or similar to the embodiment shown in FIG. 4 , so they are not repeated here, and are incorporated herein by reference.

In some embodiments, the device is further configured to: for the image recognition service request, if the target business service corresponding to the image recognition service request has been deployed on multiple container sets, use the container management platform according to the requirements. load information of each container set in the multiple container sets, determine a target container set from the multiple container sets, and process the image recognition service request through the target business service deployed on the target container set. Here, the related operations are the same as or similar to the embodiment shown in FIG. 4 , so they are not repeated here, and are incorporated herein by reference.

In some embodiments, the business service in the image recognition business service further includes an image recognition management service, and the image recognition management service is used to manage the at least one recognition service; wherein, the device is further configured to: for all Each recognition service in the at least one recognition service, when the recognition service is started, according to the domain name information and/or address information of the image recognition management service, send registration information to the image recognition management service; The recognition management service receives the registration information, generates relevant information of the recognition service according to the registration information, and stores the relevant information of the recognition service in the database service in the basic service corresponding to the image recognition service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 4 , so they are not repeated here, and are incorporated herein by reference.

In some embodiments, the device is further configured to: for each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service; if The image recognition management service does not receive the heartbeat information sent by the ninth recognition service in the at least one recognition service within a predetermined time interval, then determines the tenth recognition service in the at least one recognition service, and obtains the information from the database. The service obtains currently loaded atlas information corresponding to the ninth identification service, so as to load the atlas corresponding to the currently loaded atlas information in the tenth identification service. Here, the related operations are the same as or similar to the embodiment shown in FIG. 4 , so they are not repeated here, and are incorporated herein by reference.

8 illustrates an exemplary system that may be used to implement various embodiments described in this application.

In some embodiments, as shown in FIG. 8, system 300 can function as any of the devices in each of the described embodiments. In some embodiments, system 300 may include one or more computer-readable media (eg, system memory or NVM/storage device 320 ) having instructions and be coupled to the one or more computer-readable media and configured to execute Instructions to implement a module to perform one or more processors (eg, processor(s) 305 ) to perform the actions described herein.

For one embodiment, the system control module 310 may include any suitable interface controller to provide at least one of the processor(s) 305 and/or any suitable device or component in communication with the system control module 310 any appropriate interface.

The system control module 310 may include a memory controller module 330 to provide an interface to the system memory 315 . The memory controller module 330 may be a hardware module, a software module, and/or a firmware module.

System memory 315 may be used, for example, to load and store data and/or instructions for system 300 . For one embodiment, system memory 315 may include any suitable volatile memory, eg, suitable DRAM. In some embodiments, system memory 315 may include double data rate type quad synchronous dynamic random access memory (DDR4 SDRAM).

For one embodiment, system control module 310 may include one or more input/output (I/O) controllers to provide interfaces to NVM/storage device 320 and communication interface(s) 325 .

For example, NVM/storage device 320 may be used to store data and/or instructions. NVM/storage device 320 may include any suitable non-volatile memory (eg, flash memory) and/or may include any suitable non-volatile storage device(s) (eg, one or more hard drives ( HDD), one or more compact disc (CD) drives and/or one or more digital versatile disc (DVD) drives).

NVM/storage device 320 may include storage resources that are physically part of the device on which system 300 is installed, or it may be accessed by the device without necessarily being part of the device. For example, the NVM/storage device 320 is accessible via the communication interface(s) 325 over a network.

Communication interface(s) 325 may provide an interface for system 300 to communicate over one or more networks and/or with any other suitable device. System 300 may wirelessly communicate with one or more components of a wireless network in accordance with any of one or more wireless network standards and/or protocols.

For one embodiment, at least one of the processor(s) 305 may be packaged with the logic of one or more controllers of the system control module 310 (eg, the memory controller module 330 ). For one embodiment, at least one of the processor(s) 305 may be packaged with logic of one or more controllers of the system control module 310 to form a system-in-package (SiP). For one embodiment, at least one of the processor(s) 305 may be integrated on the same die with the logic of one or more controllers of the system control module 310 . For one embodiment, at least one of the processor(s) 305 may be integrated on the same die with logic of one or more controllers of the system control module 310 to form a system on a chip (SoC).

In various embodiments, system 300 may be, but is not limited to, a server, workstation, desktop computing device, or mobile computing device (eg, laptop computing device, handheld computing device, tablet computer, netbook, etc.). In various embodiments, system 300 may have more or fewer components and/or different architectures. For example, in some embodiments, system 300 includes one or more cameras, keyboards, liquid crystal display (LCD) screens (including touchscreen displays), non-volatile memory ports, multiple antennas, graphics chips, application specific integrated circuits ( ASIC) and speakers.

The present application also provides a computer-readable storage medium, where the computer-readable storage medium stores computer code, and when the computer code is executed, the method described in any preceding item is executed.

The present application also provides a computer program product, when the computer program product is executed by a computer device, the method according to any one of the preceding items is executed.

The present application also provides a computer device, the computer device comprising:

one or more processors;

memory for storing one or more computer programs;

The one or more computer programs, when executed by the one or more processors, cause the one or more processors to implement the method of any preceding item.

It should be noted that the present application may be implemented in software and/or a combination of software and hardware, eg, an application specific integrated circuit (ASIC), a general purpose computer, or any other similar hardware device. In one embodiment, the software program of the present application may be executed by a processor to implement the steps or functions described above. Likewise, the software programs of the present application (including associated data structures) may be stored on a computer-readable recording medium, such as RAM memory, magnetic or optical drives or floppy disks, and the like. In addition, some steps or functions of the present application may be implemented in hardware, for example, as a circuit that cooperates with a processor to perform various steps or functions.

In addition, a part of the present application can be applied as a computer program product, such as computer program instructions, which when executed by a computer, through the operation of the computer, can invoke or provide methods and/or technical solutions according to the present application. Those skilled in the art should understand that the existing forms of computer program instructions in computer-readable media include but are not limited to source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include but are not limited to Limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding post-installation program. program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium that can be accessed by a computer.

Communication media includes media by which communication signals containing, for example, computer readable instructions, data structures, program modules or other data are transmitted from one system to another. Communication media may include conducted transmission media such as cables and wires (eg, fiber optic, coaxial, etc.) and wireless (unconducted transmission) media capable of propagating energy waves, such as acoustic, electromagnetic, RF, microwave, and infrared . Computer readable instructions, data structures, program modules or other data may be embodied, for example, as a modulated data signal in a wireless medium such as a carrier wave or similar mechanism such as embodied as part of spread spectrum technology. The term "modulated data signal" refers to a signal whose one or more characteristics are altered or set in a manner that encodes information in the signal. Modulation can be analog, digital or hybrid modulation techniques.

By way of example and not limitation, computer-readable storage media may include volatile and non-volatile, readable storage media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data. Removable and non-removable media. For example, computer-readable storage media include, but are not limited to, volatile memory, such as random access memory (RAM, DRAM, SRAM); and non-volatile memory, such as flash memory, various read-only memories (ROM, PROM, EPROM) , EEPROM), magnetic and ferromagnetic/ferroelectric memory (MRAM, FeRAM); and magnetic and optical storage devices (hard disks, tapes, CDs, DVDs); or other media now known or later developed capable of storing data for computer systems Computer readable information/data used.

Here, an embodiment according to the present application includes an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein, when the computer program instructions are executed by the processor, a trigger is The apparatus operates based on the aforementioned methods and/or technical solutions according to various embodiments of the present application.

It will be apparent to those skilled in the art that the present application is not limited to the details of the above-described exemplary embodiments, but that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Accordingly, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the application is to be defined by the appended claims rather than the foregoing description, which is therefore intended to fall within the scope of the claims. All changes within the meaning and scope of the equivalents of , are included in this application. Any reference signs in the claims shall not be construed as limiting the involved claim. Furthermore, it is clear that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. Several units or means recited in the device claims can also be realized by one unit or means by means of software or hardware. The terms first, second, etc. are used to denote names and do not denote any particular order.

Claims

A method for deploying an image recognition service on a container cloud, wherein the method includes:

After the basic service in the image recognition service is deployed, a deployment file corresponding to the image recognition business service in the image recognition service is obtained, wherein the deployment file includes a service corresponding to each business service in the image recognition business service Name, image file address, number of container sets, configuration service address, the business services in the image recognition business service include at least one of the following: image recognition management service, front-end resource service, indexing service, at least one identification service, feature extraction service , account service, content management service, publishing service, image recognition API service, resource storage service;

For each business service in the image recognition business service, create a container set on the container cloud according to the deployment file and start the business service, and load the corresponding configuration file according to the configuration service address corresponding to the business service through the container set , and manage the business service through the container management platform, wherein the configuration file contains the mapping relationship between the service names and service addresses of other business services that the business service depends on, so that the business service can be The mapping relationship calls other business services that the business service depends on;

Wherein, the image recognition business service deployed through the container cloud performs at least any one of the following:

Receive an image publishing service request for a first identification map through the publishing service, obtain the first identification map, and provide the first identification map to the indexing service; obtain the first identification map through the indexing service identification map, generate a first index file corresponding to the first identification map, and return the first storage address corresponding to the first index file to the publishing service; obtain the first storage address through the publishing service, providing the first storage address to a first identification service corresponding to a first atlas corresponding to the first identification map in the at least one identification service; obtaining the first storage through the first identification service address, load the first index file according to the first storage address; or,

Receive an image recognition service request for a picture to be recognized sent by a target application through the image recognition API service, and obtain a user atlas corresponding to the to-be-recognized picture; use the image recognition API service from the database in the basic service The service obtains the second identification service corresponding to the user atlas in the at least one identification service, and provides the to-be-identified picture and the second identification service to the feature extraction service; obtains the feature extraction service through the feature extraction service. the picture to be identified and the second identification service, extract the feature information of the picture to be identified, and provide the feature information to the second identification service; obtain the feature information through the second identification service, Match the feature information with the index file currently loaded by the second identification service; if the matching is successful, obtain a second identification corresponding to the second index file that matches the feature information through the second identification service map information, and return the second recognition map information to the image recognition API service, obtain the second recognition map information through the image recognition API service, and obtain the second recognition map information from the basic service. the second AR material, and return the second AR material to the target application.
The method of claim 1, wherein the basic service includes at least one of the following:

configuration service;

database service;

file storage services;

cache database service;

Mirror warehouse service;

Configure the editor service.
The method according to claim 1, wherein, for each business service in the image recognition business services, creating a container set on the container cloud according to the deployment file and starting the business service includes:

For each business service in the image recognition business service, according to the number of container sets corresponding to the business service in the deployment file, create a container set corresponding to the business service and the number of container sets on the container cloud. The image file address corresponding to the business service in the deployment file is downloaded, the image file corresponding to the business service is downloaded, and the business service is started on the container set according to the image file.
The method of claim 1, wherein the image recognition management service is used to manage the at least one recognition service;

Wherein, the method also includes:

For each recognition service in the at least one recognition service, when the recognition service is started, send registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service;

The registration information is received through the image recognition management service, the relevant information of the recognition service is generated according to the registration information, and the relevant information of the recognition service is stored in the database service in the basic service.
The method of claim 4, wherein the method further comprises:

For each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service;

If the image recognition management service does not receive heartbeat information sent by a third recognition service in the at least one recognition service within a predetermined time interval, determine a fourth recognition service in the at least one recognition service, and determine a fourth recognition service from the at least one recognition service. The database service in the basic service obtains the currently loaded atlas information corresponding to the third identification service, so as to load the atlas corresponding to the currently loaded atlas information in the fourth identification service.
6. The method of claim 5, wherein said determining a fourth identification service among said at least one identification service comprises:

If the image recognition management service receives the registration information sent by the newly activated fifth recognition service in the at least one recognition service, the image recognition management service determines the fifth recognition service as the fourth recognition Serve.
6. The method of claim 5, wherein said determining a fourth identification service among said at least one identification service comprises:

Obtain currently loaded atlas information of each activated recognition service in the at least one recognition service from the database service through the image recognition management service;

A fourth recognition service is determined by the image recognition management service according to the currently loaded atlas information of each activated recognition service, wherein the fourth recognition service satisfies at least one of the following:

No atlases are currently loaded;

The least number of currently loaded atlases;

The currently loaded atlas has the fewest number of recognized images.
The method of claim 1, wherein the method further comprises:

For the image service request to be processed, if the target business service corresponding to the image service request to be processed has been deployed on multiple container sets, according to the load information of each container set in the multiple container sets, A target container set is determined from the plurality of container sets, and the to-be-processed image service request is processed through the target business service deployed on the target container set.
The method of claim 1, wherein the method further comprises:

Monitor the running status information of the created one or more container sets through the container management platform, and if the monitored load information of the target container set meets the predetermined capacity expansion condition, create one or more second target container sets, and create one or more second target container sets in the A business service corresponding to the target container set is deployed on one or more second target container sets.
The method of claim 1, wherein the method further comprises:

In response to the expansion instruction for the target container set, create one or more second target container sets through the container management platform, and deploy the service corresponding to the target container set on the one or more second target container sets Serve.
The method of claim 1, wherein the method further comprises:

If the container management platform determines that the target user atlas satisfies a predetermined concurrent loading condition, determine one or more concurrent identification services corresponding to the target user atlas;

The target user atlas is loaded separately through each of the one or more concurrent identification services.
The method according to claim 11, wherein the determining one or more concurrent identification services corresponding to the target user atlas comprises:

One or more concurrent identification services corresponding to the target user atlas are determined from the deployed identification services in the at least one identification service.
The method according to claim 11, wherein the determining one or more concurrent identification services corresponding to the target user atlas comprises:

One or more second target container sets are created, one or more sixth identification services are respectively deployed on each of the one or more second target container sets, and the one or more The sixth identification service is used as one or more concurrent identification services corresponding to the target user atlas.
The method of claim 1, wherein the method further comprises:

If the container management platform determines that the target user atlas satisfies a predetermined division condition, it divides the target user atlas into multiple subsets that do not intersect with each other, wherein each subset of the multiple subsets includes at least one identification map ; determine one or more identification services corresponding to the target user atlas;

A subset of the plurality of subsets is respectively loaded by each of the one or more identification services, wherein different identification services correspond to different subsets.
A method for image publishing based on an image recognition service deployed on a container cloud, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and the image recognition business service is managed through a container management platform , the business services in the image recognition business service include at least one of the following: front-end resource service, indexing service, at least one identification service, content management service, publishing service, for each business service in the image recognition business service, according to The deployment file creates a container set on the container cloud and starts the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the service name and service address of other business services that the business service depends on. The mapping relationship between, so that the business service calls other business services that the business service depends on according to the mapping relationship;

Wherein, the method includes:

Receive an image publishing service request for a target recognition map through the publishing service, obtain the target recognition map, and provide the target recognition map to the indexing service according to the service address of the indexing service;

Obtain the target identification map through the indexing service, generate a target index file corresponding to the target identification map, and return the target storage address corresponding to the target index file to the publishing service;

Obtain the target storage address through the publishing service, and provide the target storage address to a target recognition service corresponding to the target atlas set corresponding to the target recognition map in the at least one recognition service;

Obtain the target storage address through the target identification service, and load the target index file according to the target storage address.
The method according to claim 15, wherein the method further comprises receiving, at the publishing service, an image publishing service request for a target recognition map, obtaining the target recognition map, and publishing the target recognition map according to a service address of the indexing service. The object recognition map is provided to the indexing service, before:

Receive, through the front-end resource service, the user's operation information for the target identification map, and forward the operation information to the content management service;

Obtain the target recognition map and the target atlas corresponding to the target recognition map according to the operation information through the content management service, and update the target recognition in the database service in the basic service corresponding to the image recognition service The relevant information of the graph is generated, and an image publishing service request for the target recognition graph is generated and sent to the publishing service.
The method according to claim 15, wherein the target identification map is obtained through the indexing service, a target index file corresponding to the target identification map is generated, and a target storage address corresponding to the target index file is returned The publishing service includes:

Obtain the target recognition map through the indexing service, generate a target index file corresponding to the target recognition map, and provide the target index file to the file storage service in the basic service corresponding to the image recognition service;

Obtain the target index file through the file storage service, store and obtain the target storage address corresponding to the target index file, and return the target storage address to the index service;

Obtain the target storage address through the indexing service, and return the target storage address to the publishing service.
16. The method of claim 15, wherein the providing, through the publishing service, the target storage address to a target recognition service corresponding to a target atlas set corresponding to the target recognition map of the at least one recognition service ,include:

Obtain, through the publishing service, the publishing status indication information corresponding to the target atlas from the database service in the basic service corresponding to the image recognition service;

If the publication status indication information indicates that the target atlas has been published, obtain a target identification service corresponding to the target atlas in the at least one identification service from the database service through the publication service, and perform a target identification service according to the target atlas. The service address of the identification service provides the target storage address to the target identification service corresponding to the target atlas;

If the publication status indication information indicates that the target atlas has not been published, use the publication service to determine a target recognition service from the deployed recognition services in the at least one recognition service, and identify the service according to the service address of the target The object storage address is provided to the object identification service.
19. The method of claim 18, wherein the determining, by the publishing service, a target recognition service from deployed recognition services in the at least one recognition service comprises:

Obtain currently loaded atlas information corresponding to the deployed identification service in the at least one identification service from the database service through the publishing service;

Determine a target recognition service from the deployed recognition services according to the currently loaded atlas information by the publishing service, wherein the current loading information corresponding to the target recognition service satisfies at least one of the following:

The least number of currently loaded atlases;

The currently loaded atlas has the lowest number of recognition maps.
The method of claim 15, wherein the method further comprises:

For the image publishing service request, if the target business service corresponding to the image publishing service request has been deployed on multiple container sets, according to the load information of each container set in the multiple container sets, through the container management platform, A target container set is determined from the plurality of container sets, and the image publishing service request is processed through the target business service deployed on the target container set.
The method of claim 15, wherein the method further comprises:

Monitor the running status information of the created one or more container sets through the container management platform, and if the monitored load information of the target container set meets the predetermined capacity expansion condition, create one or more second target container sets, and create one or more second target container sets in the A business service corresponding to the target container set is deployed on one or more second target container sets.
The method of claim 15, wherein the method further comprises:

In response to the expansion instruction for the target container set, create one or more second target container sets through the container management platform, and deploy the service corresponding to the target container set on the one or more second target container sets Serve.
The method of claim 15, wherein the method further comprises:

If the container management platform determines that the target user atlas satisfies a predetermined concurrent loading condition, determine one or more concurrent identification services corresponding to the target user atlas;

The target user atlas is loaded separately through each of the one or more concurrent identification services.
The method of claim 15, wherein the method further comprises:

If the container management platform determines that the target user atlas satisfies a predetermined division condition, it divides the target user atlas into multiple subsets that do not intersect with each other, wherein each subset of the multiple subsets includes at least one identification map ; determine one or more identification services corresponding to the target user atlas;

A subset of the plurality of subsets is respectively loaded by each of the one or more identification services, wherein different identification services correspond to different subsets.
The method of claim 15, wherein the business service in the image recognition business service further comprises an image recognition management service for managing the at least one recognition service;

Wherein, the method also includes:

For each recognition service in the at least one recognition service, when the recognition service is started, send registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service;

The registration information is received through the image recognition management service, the relevant information of the recognition service is generated according to the registration information, and the relevant information of the recognition service is stored in the database service in the basic service corresponding to the image recognition service.
The method of claim 25, wherein the method further comprises:

For each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service;

If the image recognition management service does not receive the heartbeat information sent by the seventh recognition service in the at least one recognition service within a predetermined time interval, determine an eighth recognition service in the at least one recognition service, and determine the eighth recognition service from the at least one recognition service. The database service obtains the currently loaded atlas information corresponding to the seventh identification service, so as to load the atlas corresponding to the currently loaded atlas information in the eighth identification service.
An image recognition method based on an image recognition service deployed on a container cloud, wherein an image recognition business service corresponding to the image recognition service is deployed on the container cloud, and the image recognition business service is managed through a container management platform , the business services in the image recognition business service include at least one of the following: at least one recognition service, feature extraction service, image recognition API service, for each business service in the image recognition business service, according to the deployment file in the container cloud Create a container set and start the business service, the container set loads the configuration file corresponding to the business service, and the configuration file contains the mapping relationship between the service names and service addresses of other business services that the business service depends on, so that the business service invokes other business services that the business service depends on according to the mapping relationship;

Wherein, the method includes:

Receive, through the image recognition API service, an image recognition service request sent by the target application for the image to be recognized, and obtain a user atlas corresponding to the image to be recognized;

The target recognition service corresponding to the user atlas in the at least one recognition service is obtained from the database service in the basic service corresponding to the image recognition service through the image recognition API service, and the target recognition service of the service is extracted according to the feature. The service address provides the image to be recognized and the target recognition service to the feature extraction service;

Obtain the picture to be recognized and the target recognition service through the feature extraction service, extract the feature information of the picture to be recognized, and provide the feature information to the target recognition service according to the service address of the target recognition service Serve;

Obtain the feature information through the target recognition service, and match the feature information with the index file currently loaded by the target recognition service;

If the matching is successful, obtain the target recognition map information corresponding to the target index file matching the feature information through the target recognition service, and return the target recognition map information to the image recognition API service. The API service obtains the target recognition map information, obtains the target AR material corresponding to the target recognition map information from the basic service, and returns the target AR material to the target application.
The method according to claim 27, wherein the receiving, through the image recognition API service, an image recognition service request for a picture to be recognized and sent by a target application, and obtaining a user atlas corresponding to the to-be-recognized picture, comprises:

Receive, through the image recognition API service, an image recognition service request sent by a target application for a picture to be recognized, and perform access authority authentication on the target application;

If the authentication is passed, the user atlas corresponding to the picture to be recognized is obtained.
The method of claim 27, wherein the method further comprises:

For the image recognition service request, if the target business service corresponding to the image recognition service request has been deployed on multiple container sets, according to the load information of each container set in the multiple container sets, A target container set is determined from the plurality of container sets, and the image recognition service request is processed through the target business service deployed on the target container set.
The method of claim 27, wherein the method further comprises:

Monitor the running status information of the created one or more container sets through the container management platform, and if the monitored load information of the target container set meets the predetermined expansion conditions, create one or more second target container sets, and create one or more second target container sets in the The business service corresponding to the target container set is deployed on the one or more second target container sets.
The method of claim 27, wherein the method further comprises:

In response to the expansion instruction for the target container set, create one or more second target container sets through the container management platform, and deploy the service corresponding to the target container set on the one or more second target container sets Serve.
The method of claim 27, wherein the business service in the image recognition business service further comprises an image recognition management service for managing the at least one recognition service;

Wherein, the method also includes:

For each recognition service in the at least one recognition service, when the recognition service is started, send registration information to the image recognition management service according to the domain name information and/or address information of the image recognition management service;

The registration information is received through the image recognition management service, the relevant information of the recognition service is generated according to the registration information, and the relevant information of the recognition service is stored in the database service in the basic service corresponding to the image recognition service.
The method of claim 32, wherein the method further comprises:

For each activated recognition service in the at least one recognition service, periodically send heartbeat information to the image recognition management service through the activated recognition service;

If the image recognition management service does not receive the heartbeat information sent by the ninth recognition service in the at least one recognition service within a predetermined time interval, determine the tenth recognition service in the at least one recognition service, and determine the tenth recognition service from the at least one recognition service. The database service obtains the currently loaded atlas information corresponding to the ninth identification service, so as to load the atlas corresponding to the currently loaded atlas information in the tenth identification service.
A device for deploying an image recognition service on a container cloud, wherein the device includes:

processor; and

a memory arranged to store computer-executable instructions which, when executed, cause the processor to perform a method as claimed in any one of claims 1 to 14.
A device for image publishing based on an image recognition service deployed on a container cloud, wherein the device includes:

processor; and

a memory arranged to store computer-executable instructions which, when executed, cause the processor to perform a method as claimed in any one of claims 15 to 26.
An image recognition device based on an image recognition service deployed on a container cloud, wherein the device includes:

processor; and

a memory arranged to store computer-executable instructions which, when executed, cause the processor to perform a method as claimed in any one of claims 27 to 33.
A computer-readable medium storing instructions that, when executed, cause a system to operate the method of any one of claims 1 to 33.