CN113076964B

CN113076964B - Method and device for identifying similar objects and electronic equipment

Info

Publication number: CN113076964B
Application number: CN202010006966.1A
Authority: CN
Inventors: 申小伟; 虞新阳; 孔明明
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2020-01-03
Filing date: 2020-01-03
Publication date: 2024-01-05
Anticipated expiration: 2040-01-03
Also published as: CN113076964A

Abstract

The invention discloses a method and a device for identifying similar objects and electronic equipment, wherein the method comprises the following steps: acquiring a target object and an object set to be identified corresponding to the target object; respectively carrying out similar recognition in the object set aiming at the target object through each recognition model in at least two recognition models to obtain a list of similar objects respectively recognized; and obtaining a similar object list of the target object according to the respectively identified similar object list.

Description

Method and device for identifying similar objects and electronic equipment

Technical Field

The present invention relates to data processing technology, and more particularly, to a method of identifying similar objects, a method of identifying similar goods, an apparatus for identifying similar objects, an electronic device, and a computer-readable storage medium.

Background

At present, similar recognition is applied to many occasions, wherein the similar recognition is to recognize whether two objects are similar or not, for example, distribution errors are avoided through similar recognition in logistics distribution, distribution accuracy is improved, and for example, an e-commerce platform recommends similar commodities of selected commodities for a user through similar recognition of the commodities selected by the user, so that shopping experience of the user is improved, and the like.

In the prior art, the similarity recognition is to determine whether two objects are similar through one recognition model, for example, determine whether images of two objects are similar through one deep learning model, and the like. In the similarity recognition by the recognition model, a similarity threshold needs to be set to divide similar objects and dissimilar objects, and here, since the similarity threshold is usually an empirical value, but the similarity threshold plays a decisive role in recognizing whether two objects are similar, the similarity recognition by the recognition model in the prior art has a problem of lower accuracy, which is shown, for example, in that a similar object list obtained by recognizing any object is incomplete, or a similar object list obtained by recognizing any object contains dissimilar objects.

Disclosure of Invention

An object of the embodiment of the invention is to provide a new technical scheme for performing similar recognition so as to improve accuracy of similar recognition.

According to a first aspect of the present invention there is provided a method of identifying similar objects comprising:

acquiring a target object and an object set to be identified corresponding to the target object;

respectively carrying out similar recognition in the object set aiming at the target object through each recognition model in at least two recognition models to obtain a list of similar objects respectively recognized;

And obtaining a similar object list of the target object according to the respectively identified similar object list.

Optionally, the performing similar recognition on the target object in the object set through each recognition model of at least two recognition models, and obtaining the list of similar objects respectively recognized includes:

obtaining a similarity index value between each object in the object set and the target object through each identification model;

and for each recognition model, obtaining a similar object list recognized by the corresponding recognition model according to the similar index value obtained by the corresponding recognition model.

Optionally, the obtaining the list of similar objects identified by the corresponding identification model according to the similarity index value obtained by the corresponding identification model includes:

and selecting a set number k objects with the highest similarity index value from the object set according to the similarity index value obtained by the corresponding recognition model, and forming a similarity object list recognized by the corresponding recognition model.

Optionally, the method further comprises a step of determining the set number k, including:

obtaining a current similar object list respectively identified by each identification model according to the current numerical value of the set number k;

Acquiring a next similar object list respectively identified by each identification model when the set number k is a next numerical value larger than the current numerical value;

comparing the corresponding next similar object list with the corresponding current similar object list for each identification model to obtain a corresponding newly added object list;

determining that the current numerical value is the final value of the set number k under the condition that the occurrence frequencies of at least part of objects in the current similar object list in all the newly added object lists are smaller than a set first frequency threshold value;

and adjusting the current value of the set value under the condition that the occurrence frequency of any object in the at least part of objects in all the newly added object lists is greater than or equal to the first frequency threshold value.

Optionally, the step of determining the set number k further includes a step of selecting the at least part of the objects, including:

and selecting the objects with the occurrence frequency greater than or equal to a set second frequency threshold value from all the current similar object lists, and selecting at least part of the objects as the objects.

Optionally, the step of determining the set number k further includes:

And increasing the value of the set number k according to the set step distance on the basis of the current value of the set number k as the next value.

Optionally, the adjusting the current value of the set value includes:

updating the current value of the set value to be equal to the next value.

Optionally, the obtaining a similarity index value between each object in the object set and the target object includes:

extracting vector values of the feature vectors set by the target object for the corresponding recognition model through the corresponding recognition model, and extracting vector values of each object in the object set for the feature vectors;

for each object in the object set, a distance value between the vector value of the corresponding object and the vector value of the target object is obtained through a corresponding identification model and is used as a similarity index value between the corresponding object and the target object.

Optionally, the obtaining the similar object list of the target object according to the respectively identified similar object list includes:

for each object in all the similar object lists, obtaining a similarity score of the corresponding object according to the occurrence frequency of the corresponding object in all the similar object lists;

And screening objects similar to the target object from all the similar object lists according to the similarity scores of each object in all the similar object lists, and forming a similar object list of the target object.

Optionally, the screening the objects similar to the target object in all the similar object lists according to the similarity scores of each object in all the similar object lists includes:

and screening objects with the similarity score being greater than or equal to a set score threshold value from all the similar object lists as objects similar to the target object.

Optionally, the obtaining the target object and the object set to be identified corresponding to the target object includes:

in response to a selection operation for a plurality of objects, selecting any object from the selected objects as the target object, and selecting other objects to form an object set to be identified corresponding to the target object.

in response to a selection operation for any object in the object list, selecting the selected object as a target object, and selecting other objects in the object list to form an object set corresponding to the target object.

Optionally, the method further comprises:

and outputting a similar object list of the target object.

Optionally, the outputting the similar object list of the target object includes:

and sending the similar object list of the target object to terminal equipment bound with a target account, wherein the target account is an account for providing distribution service for the target object and the objects in the object set.

Optionally, the method further comprises:

and selecting at least two recognition models matched with the object type from a recognition model set according to the object type of the target object.

According to a second aspect of the present invention there is also provided a method of identifying similar items comprising:

acquiring a commodity set of commodities with at least one same distribution characteristic;

for each commodity in the commodity set, respectively carrying out similar recognition on the corresponding commodity in the commodity set through each recognition model in at least two recognition models to obtain a similar commodity list which is recognized respectively;

and obtaining a similar commodity list of the corresponding commodity according to the respectively identified similar commodity list.

Optionally, the acquiring the commodity set of commodities having at least one same distribution characteristic includes:

And acquiring a commodity set of commodities corresponding to the same distribution box and the same distribution time.

Optionally, the method further comprises:

and sending the obtained similar commodity list to terminal equipment bound with a target account, wherein the target account is a distributor account for providing distribution service for a corresponding distribution box at corresponding distribution time.

According to a third aspect of the present invention there is also provided a method of identifying similar items comprising:

for each commodity in the commodity set, obtaining a similar list of the corresponding commodity in the commodity set according to a similar commodity list of the corresponding commodity in a commodity library, which is obtained in advance;

providing the obtained similarity list through the terminal equipment bound with the target account;

wherein obtaining a list of similar commodities corresponding to the commodity in the commodity library comprises:

respectively carrying out similar recognition on corresponding commodities in the commodity library through each recognition model in at least two recognition models to obtain a similar commodity list which is respectively recognized;

and obtaining a similar commodity list of the corresponding commodity in the commodity library according to the respectively identified similar commodity list.

According to a fourth aspect of the present invention there is also provided a method of identifying similar items comprising:

acquiring a commodity selected in a current commodity list as a target commodity;

respectively carrying out similar recognition in the current commodity list aiming at the target commodity through each recognition model in at least two recognition models to obtain a similar commodity list which is respectively recognized;

obtaining a similar commodity list of the target commodity according to the respectively identified similar commodity list;

and displaying the similar commodity list in the current commodity list.

According to a fifth aspect of the present invention, there is also provided a similarity identifying method comprising:

obtaining a similar list of the target commodity in a current commodity list according to a similar commodity list of the target commodity in a commodity library, which is obtained in advance;

displaying the similar commodity list in the current commodity list;

wherein obtaining a list of similar articles of the target article in the article library comprises:

respectively carrying out similar recognition on the target commodity in the commodity library through each recognition model in at least two recognition models to obtain a similar commodity list which is respectively recognized;

And obtaining a similar commodity list of the target commodity in the commodity library according to the respectively identified similar commodity list.

According to a sixth aspect of the present invention, there is also provided an apparatus for identifying similar objects, comprising:

the object acquisition module is used for acquiring a target object and an object set to be identified corresponding to the target object;

the recognition processing module is used for carrying out similar recognition on the target object in the object set through each recognition model in at least two recognition models to obtain a list of similar objects recognized by the recognition processing module; the method comprises the steps of,

and the list generation module is used for obtaining a similar object list of the target object according to the respectively identified similar object list.

According to a seventh aspect of the present invention there is also provided an electronic device comprising an apparatus according to the third aspect of the present invention; or,

the electronic device includes a memory and a processor;

the memory stores a computer program which, when executed by the processor, implements the method according to any one of the first to fifth aspects of the invention.

According to a fifth aspect of the present invention there is also provided a computer readable storage medium storing a computer program readable by a computer for performing the method according to any one of the first to fifth aspects of the present invention when the computer program is read and run by the computer.

Based on the similar recognition scheme provided by the embodiment of the invention, the target object can be subjected to similar recognition through at least two recognition models, so that a similar object list which is respectively recognized through each recognition model is obtained, objects similar to the target object are screened out according to the obtained similar object lists, and a similar object list of the target object is generated. According to the similarity recognition scheme, objects similar to the target object can be screened according to the similarity object lists respectively recognized by the recognition models, so that the influence of a similarity threshold on a similarity recognition result can be reduced, and the accuracy of similarity recognition is improved.

Other features of the present invention and its advantages will become apparent from the following detailed description of exemplary embodiments of the invention, which proceeds with reference to the accompanying drawings.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description, serve to explain the principles of the invention.

FIG. 1a is a schematic illustration of an alternative application scenario of an embodiment of the present invention;

FIG. 1b is a schematic illustration of an alternative application scenario according to an embodiment of the present invention;

FIG. 2 is a schematic block diagram of a hardware architecture of an electronic device that can be used to perform a method of identifying similar objects in accordance with an embodiment of the invention;

FIG. 3 is a flow diagram of a method of identifying similar objects, according to one embodiment;

FIG. 4 is a flow chart of the steps of determining a set number k according to one embodiment;

FIG. 5 is a schematic diagram of the operational steps for determining a set number k according to one example;

FIG. 6 is a flow diagram of a method of identifying similar items according to one embodiment;

FIG. 7 is a flow chart of a method of identifying similar items according to yet another embodiment;

FIG. 8 is a flow diagram of an apparatus for identifying similar objects, according to one embodiment.

Detailed Description

Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that: the relative arrangement of the components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present invention unless it is specifically stated otherwise.

The following description of at least one exemplary embodiment is merely exemplary in nature and is in no way intended to limit the invention, its application, or uses.

Techniques, methods, and apparatus known to one of ordinary skill in the relevant art may not be discussed in detail, but are intended to be part of the specification where appropriate.

In all examples shown and discussed herein, any specific values should be construed as merely illustrative, and not a limitation. Thus, other examples of exemplary embodiments may have different values.

It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further discussion thereof is necessary in subsequent figures.

The method of the present embodiment may be used to identify similar objects of the target object to generate a similar object list of the target object, where the object may be any article, a face image, and the like, and is not limited herein. According to the method, at least two recognition models are used for carrying out similar recognition on the target object respectively to obtain respective similar object lists, and similar objects of the target object are screened out from the similar object lists in a voting mode and the like according to the similar object lists obtained through the recognition models, so that a similar object list of the target object is generated. The embodiment may be applied to any scene that needs to be similarly identified, such as a logistics distribution scene, a similar commodity pushing scene, a person tracking scene, etc., that is, the object in the embodiment may be a commodity, other articles, a face image, etc., which is not limited herein.

Fig. 1a shows an application of the method of the present embodiment in a logistics distribution scenario, in which a distribution system distributes corresponding commodities to each distribution box according to an order address or the like, for example, distributes commodity 1, commodity 2, commodity 3, commodity 4, commodity 5, and the like to a 1# distribution box for distribution, and distributes commodity 1, commodity 5, commodity 6, commodity 7, commodity 8 to a 1# distribution box for distribution, and the like. In the logistics distribution scene, if similar commodities exist in the same distribution box, a distribution operator is easy to generate a distribution error in the process of distributing the commodities, so that for each distribution box, similar identification can be performed on each commodity in the corresponding distribution box in the commodity range of the corresponding distribution box through the similar identification method of the embodiment so as to obtain a similar commodity list of each commodity, and the similar commodity list is sent to a mobile phone of the distribution operator of the corresponding distribution box so as to remind the distribution operator of the similar commodity in the distribution box in charge of the distribution operator, and the distribution error is avoided.

As shown in fig. 1a, according to the method of the present embodiment, for example, three different recognition models are provided, namely, a recognition model M1, a recognition model M2, and a recognition model M3. Taking the 1# distribution box as an example, according to the method of the present embodiment, for each product in the 1# distribution box, a respective similar product list may be identified within the product range of the 1# distribution box. Taking the commodity 1 in the 1# distribution box as an example, taking the commodity 1 as a target commodity, taking other commodities in the 1# distribution box as a commodity set to be identified of the corresponding commodity 1, identifying the commodity similar to the commodity 1 in the other commodities in the 1# distribution box through an identification model M1, and forming a similar commodity list S1 of the commodity 1; identifying commodities similar to the commodity 1 in other commodities in the 1# distribution box through the identification model M2, and forming a similar commodity list S2 of the commodity 1; identifying the commodities similar to the commodity 1 in other commodities in the 1# distribution box through the identification model M2, and forming a similar commodity list S3 of the commodity 1; finally, a list of similar items of the item 1 in the 1# distribution box is obtained according to the similar item list S1, the similar item list S2 and the similar item list S3, which may be obtained by voting for the items in the similar item lists S1, S2 and S3. After identifying the similar product list of the corresponding product in the product range of the 1# distribution box for each product in the 1# distribution box, the similar product list can be sent to the mobile phone 3100A of the distributor of the 1# distribution box, so that the distributor can identify the similar product in the 1# distribution box in a key manner when distributing the product by inquiring the similar product list of each product in the 1# distribution box on the mobile phone 3100A, and further, delivery errors are avoided.

Similarly, taking the 2# distribution box as an example, according to the method of the present embodiment, for each product in the 2# distribution box, a respective similar product list may be identified within the product range of the 2# distribution box. For example, the 2# distribution box also has the commodity 1, taking the commodity 1 in the 2# distribution box as a target commodity, taking other commodities in the 2# distribution box as a commodity set to be identified of the corresponding commodity 1, identifying the commodity similar to the commodity 1 in the other commodities in the 2# distribution box through the identification model M1, and forming a similar commodity list S4 of the commodity 1; identifying commodities similar to the commodity 1 in other commodities in the 2# distribution box through the identification model M2, and forming a similar commodity list S5 of the commodity 1; identifying the commodities similar to the commodity 1 in other commodities in the 2# distribution box through the identification model M3, and forming a similar commodity list S6 of the commodity 1; finally, a similar item list of the commodity 1 in the 2# distribution box is obtained according to the similar commodity list S4, the similar commodity list S5 and the similar commodity list S6. After identifying the similar product list of the corresponding product in the product range of the 2# distribution box for each product in the 2# distribution box, the similar product list can be sent to the mobile phone 3100B of the distributor of the 2# distribution box, so that the distributor can identify the similar product in the 2# distribution box in a key manner when distributing the product by inquiring the similar product list of each product in the 2# distribution box on the mobile phone 3100B, and further, delivery errors are avoided.

Fig. 1b shows an application of the method of the present embodiment in a merchandise push scenario. In this scenario, for example, the user searches for "sports pants" in the client of the e-commerce platform through the user terminal 2000, the client will provide a commodity list with a commodity label of "sports pants", on the basis that, if the user selects the commodity 1 in the commodity list, for example, clicks on the list item of the commodity 1, the user may select the selected commodity 1 as a target commodity, and select other commodities in the commodity list to form a commodity set to be identified of the selected commodity 1, so as to apply the method according to the embodiment, identify a commodity similar to the selected commodity 1 in the commodity set, and further obtain a similar commodity list of the selected commodity 1, and recommend the similar commodity list to the user, so as to reduce the screening range of the user when screening the required commodity. In this scenario, according to the method of the embodiment, the similar commodity of the commodity 1 is identified in the searched other commodities by the identification model M1, so as to form a similar commodity list S7 of the commodity 1; identifying commodities similar to the commodity 1 in the other searched commodities through the identification model M2 to form a similar commodity list S8 of the commodity 1; and identifying the commodities similar to the commodity 1 in the other searched commodities through the identification model M3 to form a similar commodity list S9 of the commodity 1; finally, according to the similar commodity list S7, the similar commodity list S8 and the similar commodity list S9, a similar commodity list of the commodities 1 in the searched commodities is obtained, wherein the similar commodity list comprises, for example, commodities 1-1, commodities 1-2, commodities 1-3, commodities 1-4 and the like, similar commodity items are added in the commodity list based on the similar commodity list, and list items of the commodities in the similar commodity list are provided under the similar commodity items so that a user can quickly select.

For any application scenario of the similar recognition method of the present embodiment, the method for recognizing the similar object may be implemented online in each event requiring the similar recognition method, and hereinafter, the method for recognizing the similar object, the method for recognizing the similar commodity, and the like will be collectively referred to simply as the similar recognition method.

For any application scenario of the similarity recognition method of the present embodiment, in the background, similarity recognition may be performed on all objects in the object library in advance according to the similarity recognition method of the present embodiment, for example, on all commodities in the commodity library, so as to generate a similarity list of each object in the object library, and then, when the above event occurs, the similarity list of the target object in the object library may be obtained and output by searching the similarity list of the target object in the object library. For example, if the set of objects to be identified in the corresponding event includes the object 1 to the object 4, for each object in the object 1 to the object 4, whether the corresponding object exists or not may be searched in a similarity list of the target object in the object library, and if the corresponding object exists, it is indicated that the similarity list of the target object in the corresponding event includes the corresponding object.

< hardware configuration >

Fig. 2 shows a schematic block diagram of a hardware architecture of an electronic device that can be used to perform the similar identification method of any of the embodiments of the invention.

As shown in fig. 2, the electronic device 1000 may be any type of terminal device, or may be any type of server, including a cloud server, a server cluster, and the like, which is not limited herein.

The terminal device may be a portable computer, a desktop computer, a mobile phone, a tablet computer, etc., which is not limited herein.

As shown in fig. 2, the electronic device 1000 may include a processor 1100, and a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, and the like, all of which are connected to the processor. Wherein the processor 1100 is adapted to execute a computer program. The computer program may be written in an instruction set of an architecture such as x86, arm, RISC, MIPS, SSE, etc. The memory 1200 includes, for example, ROM (read only memory), RAM (random access memory), nonvolatile memory such as a hard disk, and the like. The interface device 1300 includes, for example, a USB interface, a headphone interface, and the like. The communication device 1400 can be capable of wired or wireless communication, and specifically can include Wifi communication, bluetooth communication, 2G/3G/4G/5G communication, and the like. The display device 1500 is, for example, a liquid crystal display. The input device 1600 may include, for example, a touch screen, keyboard, voice input, somatosensory input, and the like. The speaker 1700 is for outputting audio signals. Microphone 1800 is used to collect audio signals.

The memory 1200 of the electronic device 1000 is used for storing instructions (computer program) for controlling the processor 1100 to operate to perform the similar recognition method provided according to any of the embodiments of the present invention. The skilled person can design instructions according to the disclosed solution. How the instructions control the processor to operate is well known in the art and will not be described in detail here.

It will be appreciated by those skilled in the art that although a plurality of devices are shown for electronic device 1000 in fig. 1, the present invention may relate to only some of the devices, e.g., electronic device 1000 may relate to only processor 1100 and memory 1200, etc.

In further embodiments, the electronic device 1000 may further comprise a similar recognition means according to any embodiment of the invention, the respective modules of which may be implemented by the processor 1100 of the electronic device 1000 in the above embodiments.

< method example one >

FIG. 3 is a flow diagram of a method of similarity identification, according to one embodiment. In this embodiment, the similarity identifying method may be implemented by any electronic device, for example, the electronic device 1000 in fig. 1, which may be a terminal device or a server.

As shown in fig. 3, the similar identification method of the present embodiment may include the following steps S3100 to S3300:

step S3100, a target object and a set of objects to be identified corresponding to the target object are obtained.

In this embodiment, the object may be any entity, for example, an article such as a commodity, any object, a face image, or the like, which is not limited herein.

In this embodiment, an object similar to the target object needs to be identified in the object set to be identified, so as to obtain a similar object list of the target object, that is, the similar object list includes objects similar to the target object.

In one embodiment, the method is applied to identify, among a selected plurality of objects, any scene of other objects similar to any object therein, for example, to a logistics distribution scene as shown in FIG. 1 a.

In this embodiment, the obtaining the target object and the object set to be identified corresponding to the target object in step S3100 may include: in response to a selection operation for a plurality of objects, selecting any object from the selected plurality of objects as the target object, and selecting other objects to form an object set to be identified corresponding to the target object.

In an embodiment, the method may also be applied to any scenario in which for a selected one of the objects, other objects similar to the object are identified, e.g. to a merchandise push scenario as shown in fig. 1b, etc.

In this embodiment, the obtaining the target object and the object set to be identified corresponding to the target object in step S3100 may include: in response to a selection operation for any object in the object list, the selected object is selected as a target object, and other objects in the object list are selected to form an object set corresponding to the target object.

In this embodiment, the object list is used, for example, to show list items of objects queried according to the search conditions entered by the user, i.e. the object list contains list items of individual objects queried according to the search conditions. The list item may contain basic information of the corresponding object, etc., and by clicking on the list item, a detail interface of the corresponding object may be entered.

Step S3200, performing similar recognition in the object set for the target object through each recognition model in at least two recognition models to obtain a list of similar objects respectively recognized.

In this embodiment, two different recognition models or two or more different recognition models may be provided.

The recognition model may be any model capable of performing similar recognition.

The recognition model may include two parts, one being a feature extraction model and the other being a similar calculation model. The feature extraction model is used to extract a vector value of an arbitrary object with respect to a set feature vector, which may be one-dimensional, i.e., composed of one feature, or multidimensional, i.e., composed of a plurality of features, to represent a corresponding object by the vector value, which is composed of a feature value of an arbitrary object with respect to each feature. For example, a feature vector includes 15 features, and then any object includes 15 feature values for the vector value of the feature vector.

The similarity calculation model is used for calculating a similarity index value representing the degree of similarity between two objects according to the vector values of the two objects, the similarity index value being, for example, a distance value representing the distance between the two vector values, wherein the larger the similarity index value, the more similar the two objects are.

The feature extraction model may be based on any deep learning model. The similarity index value calculated by the similarity calculation model may be, for example, an euclidean distance value, a manhattan distance value, a mahalanobis distance value, a cosine distance value, a pearson correlation coefficient, or the like, and is not limited thereto.

For the two recognition models, as long as any one of the feature extraction model and the similar calculation model is different, the two recognition models are different recognition models.

The at least two recognition models in this embodiment may include, for example, at least one model of a Siamese model, a deep ranking model, and an acceptance model.

According to step S3200, objects similar to the target object may be screened out from the object set by each of the at least two recognition models, so as to obtain a list of similar objects identified by the at least two recognition models, i.e. the list of similar objects includes objects similar to the target object screened out from the object set by the corresponding recognition model.

Still taking the scenario of fig. 1 as an example, for the example of the three recognition models M1, M2, M3 provided, according to step S3200, a list of similar objects recognized by the recognition model M1, a list of similar objects recognized by the recognition model M2, and a list of similar objects recognized by the recognition model M3 will be obtained.

In this example, since the three recognition models M1, M2, M3 differ in at least one of feature extraction and similarity calculation, there is a high possibility that the objects included in the respective recognized similar object lists differ.

In one embodiment, for any recognition model, a similarity index value of each object in the object set for the target object is obtained through the recognition model, and a similar object list obtained through recognition of the recognition model is obtained according to the obtained similarity index value, which is beneficial to consuming less data processing amount and obtaining the similar object list obtained through recognition of the recognition model.

In this embodiment, in step S3200, through each recognition model of at least two recognition models, similar recognition is performed on the target object in the object set to obtain a list of similar objects identified respectively, which may include steps S3210 to S3220 as follows:

step S3210, obtaining a similarity index value between each object in the object set and the target object through each recognition model.

In step S3210, each recognition model is called to obtain a similarity index value between each object in the object set and the target object.

Continuing with fig. 1 as an example, in step S3210, a similarity index value between each object in the set of objects and the target object is obtained by the recognition model M1, that is, one object corresponds to one similarity index value obtained by the recognition model M1; obtaining a similarity index value between each object in the object set and the target object through the recognition model M2, namely, obtaining a similarity index value by the recognition model M2 corresponding to one object; and obtaining a similarity index value between each object in the object set and the target object by the recognition model M3, that is, one object corresponds to one similarity index value obtained by the recognition model M3.

In the step S3210, obtaining a similarity index value between each object in the object set and the target object may include:

in step S3211, the vector value of the feature vector set by the target object for the corresponding recognition model is extracted through the corresponding recognition model, and the vector value of each object in the object set for the feature vector is extracted.

In this step S3211, for each recognition model, a feature vector whose feature extraction model needs to be extracted may be set in advance.

Different recognition models may set different feature vectors, or the same feature vector may be set, and the present invention is not limited thereto.

The recognition model may extract vector values of feature vectors set by the object for the recognition model from the object data. The object data includes at least one of an object image, text information describing the object.

Step S3212, for each object in the object set, obtaining, by the corresponding recognition model, a distance value between the vector value of the corresponding object and the vector value of the target object as a similarity index value between the corresponding object and the target object.

Through the above steps S3211 and S3212, the similarity index value between each object in the object set and the target object may be obtained through each recognition model, so as to obtain a list of similar objects recognized by the corresponding recognition model according to the ranking of the similarity index values.

S3220, for each recognition model, obtaining a list of similar objects recognized by the corresponding recognition model according to the similarity index values obtained by the corresponding recognition model.

In this embodiment, for each recognition model, a set number k of objects that make the similarity index value highest may be selected from the object set according to the order of the similarity index values obtained by the corresponding recognition model, to form a list of similar objects recognized by the corresponding recognition model.

The different recognition models may correspond to the same set number k, or may correspond to different set numbers k, which is not limited herein.

With continued reference to fig. 1, for example, the set number k of each recognition model is set to be 5, and the set object set includes 30 objects, then the similarity index value of each object of the 30 objects for the target object is obtained through the recognition model M1, and 5 objects with the highest similarity index value are selected from the 30 objects, so as to form a similar object list recognized by the recognition model M1; obtaining a similarity index value of each object of the 30 objects for the target object through the recognition model M2, and selecting 5 objects with highest similarity index values from the 30 objects to form a similarity object list recognized by the recognition model M2; and the recognition model M3 obtains the similarity index value of each object of the 30 objects for the target object, and selects 5 objects with the highest similarity index value from the 30 objects to form a similar object list recognized by the recognition model M3.

In one embodiment, the set number k corresponding to each recognition model may be a fixed value that is set in advance.

Because the number of similar objects in the object set of different target objects is different, a relatively accurate recognition result can be obtained by setting a fixed set number k for some target objects, and the problem that the finally obtained similar object list contains dissimilar objects or similar objects in the similar object list are not complete can possibly occur for other target objects.

In step S3300, a list of similar objects of the target object is obtained according to the list of similar objects identified by each of all the identification models.

In the step S3300, a similarity score of each object in each similar object list may be obtained according to the occurrence of each object in each similar object list, and whether the corresponding object can be regarded as an object similar to the target object may be determined according to the similarity score, and the obtained similarity score may be listed in the final similar object list.

Thus, in one embodiment, the obtaining the similar object list of the target object according to the similar object list respectively identified by all the identification models in the step S3300 may include the following steps S3310 to S3330:

in step S3310, for each object in all the similar object lists, a similarity score of the corresponding object is obtained according to the occurrence of the corresponding object in all the similar object lists.

The occurrence may include a frequency of occurrence and the similarity score may simply be equal to the frequency of occurrence of the corresponding object in each of the similar object lists. For example, if an object appears in the list of similar objects identified by the two identification models, the similarity score of the object is equal to 2.

The occurrence may also include a list of similar objects in which the corresponding object appears, and correspondingly, the similarity score may also be determined from the list of similar objects in which the corresponding object appears. Because the similar object list where the corresponding object appears has a mapping relation with the recognition models, different weights can be set for different recognition models, and the similarity scores of the corresponding objects can be calculated according to the weights. For example, if an object appears in a list of similar objects identified by two identified models, then the similarity score for that object will be equal to the sum of the weights of the two identified models, and so on.

In step S3320, according to the similarity score of each object in all the similar object lists, objects similar to the target object are screened out from all the similar object lists to form a similar object list of the target object.

According to this step, for example, objects having a similarity score greater than or equal to a set score threshold value may be screened out from all the similar object lists as objects similar to the target object, thereby forming a similar object list of the target object.

In the case where the similarity score is equal to the occurrence frequency, the score threshold may be determined according to the number of set recognition models, the score threshold may be valued as a majority of the recognition models, for example, three recognition models may be set, the score threshold may be set to 2, five recognition models may be set, the score threshold may be set to 3, or the like, and is not limited herein.

To facilitate setting the scoring threshold, in one embodiment, at least two recognition models may be set as an odd number of recognition models.

According to the above steps S3100 to S3300, the method in this embodiment performs similar recognition on the target object in the range of the corresponding object set through at least two recognition models on the respective channels, so as to obtain the similar object list of the target object identified by each recognition model, and obtain the similar object list of the target object according to the similar object lists.

In one embodiment, the set number k may be adaptively adjusted during the recognition process according to the situation of the similar object list recognized by each recognition model, so that each recognition model can obtain the similar object list recognized by each recognition model based on the adaptively adjusted set number k.

In this embodiment, as shown in fig. 4, the method of determining the value of the set number k may include the following steps S4100 to S4500:

in step S4100, a list of current similar objects identified by each of the identification models is obtained according to the current value of the set number k.

In this embodiment, an initial value of the set number k may be set in advance, and the initial value may be a smaller value, for example, the initial value is taken as 2 or 3, or the like.

In this embodiment, in the first recognition operation according to the set number k, the current value is the initial value of the set number k. In the second recognition operation according to the set number k, the current value is equal to the value after the set step is increased on the basis of the initial value.

The set stride may be determined in consideration of the similarity recognition accuracy and the processing speed, for example, the set stride is selected to be 1 so as to preferentially obtain the highest similarity recognition accuracy.

Thus, for example, if the initial value of the set number k is 2, the current value of the set number k is 3 in the second recognition operation, and then the current value of the set number increases by 1 step for each recognition operation.

In step S4200, a list of next similar objects identified by each of the identification models when the set number k is the next value greater than the current value is acquired.

The next value may be a value obtained by increasing the set number k by the set step pitch on the basis of the current value. For this, the step of determining the set number k may further include: the value of the set number k is increased as the next value according to the set step distance on the basis of the current value of the set number k.

The next value may be any other value greater than the current value, and is not limited herein.

Step S4300, for each recognition model, compares the corresponding next similar object list with the corresponding current similar object list to obtain a corresponding newly added object list.

According to this step S4300, a list of newly added objects is obtained for each recognition model.

Referring to the example of fig. 1, according to the step S4300, for the recognition model M1, a new object of the next similar object list recognized by the recognition model M1 is obtained with respect to the current similar object list recognized by the recognition model M1, and a corresponding new object list is obtained; for the recognition model M2, obtaining a new object of the next similar object list obtained through recognition of the recognition model M2 relative to the current similar object list obtained through recognition of the recognition model M2, and further obtaining a corresponding new object list; and for the recognition model M3, obtaining a new object of the next similar object list recognized by the recognition model M3 relative to the current similar object list recognized by the recognition model M3, and further obtaining a corresponding new object list. That is, in this example, three newly added object lists will be obtained through step S4300.

In step S4400, when the occurrence frequency of at least some of the objects in the current similar object list in all of the newly added object lists is less than the set first frequency threshold, the current value is determined to be the final value of the set number k.

In the step S4400, the at least part of the objects may be selected from all the current similarity list as reference objects.

For example, each object in the list of all current similar objects may be considered the reference object.

For another example, an object satisfying the set condition in all the current similar object lists may be used as the reference object.

The setting condition is, for example, that the occurrence frequency in all the current similar object lists is greater than or equal to a set second frequency threshold.

The setting condition may be, for example, the same as the scoring threshold value of the similarity score, and is not limited herein.

After the final value of the set number k is determined, the operation of determining the set number k can be ended, and the list of similar objects identified by the multiple identification models can be further processed according to the final value of the set number k.

The first frequency threshold may be determined in consideration of the similarity recognition accuracy and the processing speed, for example, the first frequency threshold may be 1 in consideration of avoiding missing similar objects as much as possible.

The first frequency threshold may also be determined according to the number of the set recognition models, for example, the frequency threshold may be a value of a plurality of the recognition models, and the like, which is not limited herein.

When the first frequency threshold is 1, according to the step S4400, as long as the occurrence frequency of each reference object in all the newly added object lists is less than 1, i.e. no reference object appears in the newly added object lists, the final value of the set number k is determined to be the current value.

In step S4500, the current value of the set value is adjusted if the occurrence frequency of any one of at least some of the objects (i.e. reference objects) in all the newly added object lists is greater than or equal to the first frequency threshold.

After adjusting the current value of the set value k, it may be determined whether the adjusted current value is the final value of the set value k through the above steps S4100 to S4500 again. The adjustment may be performed according to a set pitch, or may be performed randomly, and is not limited thereto.

The adjusting of the current value of the set value in step S4500 may include: updating the current value of the set value is equal to the next value.

When the first frequency threshold is 1, according to step S4500, if any reference object appears in any newly added object list, it is stated that, in the case where the set value k is the current value, at least one recognition model recognizes that the object similar to the target object is missing in the obtained similar object list, and the value of the set number k needs to be adjusted.

After step S4500, the process returns to step S4100 to continue step S4100 until the final value of the set number k of identification models is determined.

Fig. 5 shows an example of determining the set number k, in which example in fig. 5 three recognition models M1, M2, M3 are provided to participate in recognizing similar objects of the target object 15. In this example, the first frequency threshold is set to 1, the second frequency threshold is set to 2, the initial value of the set number k is taken to 3, the set stride is 1, the similarity score is represented by the occurrence frequency, and the score threshold is set to 2.

As shown in fig. 5, for the target object 15, the similarity index value between each object in the object set and the target object 15 is determined by the recognition model M1, and the objects in the first 8 bits in the object set include the object 01, the object 08, the object 05, the object 04, the object 02, the object 18, the object 11, and the object 10 in order of the similarity index value from high to low. For the target object 15, the recognition model M2 determines the similarity index value between each object in the object set and the target object 15, and the objects in the first 8 bits in the object set include the object 01, the object 05, the object 02, the object 08, the object 04, the object 18, the object 07 and the object 12 according to the arrangement of the similarity index values from high to low. The recognition model M3 determines the similarity index value between each object in the object set and the target object 15, and the objects in the first 8 bits in the object set include the object 01, the object 08, the object 06, the object 04, the object 05, the object 16, the object 14 and the object 13 according to the arrangement of the similarity index values from high to low.

When the set number k is adaptively adjusted to determine the final value of the set number k, a first recognition operation is performed, and at this time, the current value of the set number k is an initial value 3. According to step S4100, as shown in fig. 5, obtaining the current similar object list identified by the identification model M1 will include the object with the highest 3 similarity index values, that is, object 01, object 08, and object 05; obtaining a current similar object list identified by the identification model M2, wherein the current similar object list comprises 3 objects with highest similarity index values, namely an object 01, an object 05 and an object 02; and obtaining the current similar object list identified by the identification model M3 includes the object with the highest 3 similarity index values, namely, object 01, object 08 and object 06.

Then, the set value k is increased by the set step distance 1, and the next value of the set value k is obtained as 4.

In the case where the set number k is 4, according to step S4200, as shown in fig. 5, the next similar object list to be obtained identified by the identification model M1 includes 4 objects having the highest similarity index values, that is, object 01, object 08, object 05, and object 04; the next similar object list to be obtained, which is identified by the identification model M2, includes 4 objects having the highest similarity index values, namely, object 01, object 05, object 02, and step 08; and, the next similar object list to be obtained identified by the identification model M3 includes the object having the highest 4-similarity index value, that is, includes the object 01, the object 08, the object 06, and the object 04.

Continuing according to step S4300, a new object list that may be obtained for the corresponding recognition model M1 includes object 04, a new object list that is obtained for the corresponding recognition model M2 includes object 08, and a new object list that is obtained for the corresponding recognition model M3 includes object 04.

After each newly added object list is obtained, all the current similar object lists can be screened

At least part of the objects whose occurrence frequency is greater than or equal to the set second frequency threshold (2 in this example) are obtained as reference objects including object 01, object 08, and object 05.

After the reference object is obtained, the judgment is performed according to step S4400 and step S4500, in this example, since the reference object 08 appears 1 time in each newly added object list, and the requirement of the first frequency threshold is met, the current value of the set value needs to be updated to be 4 according to step S4500, so that the second recognition operation is performed.

When the second recognition operation is performed, the current value of the value is set to be 4, and each current similar object list obtained in step S4100 is the next similar object list in the first recognition operation, which is not described herein.

Then, the set value k is increased by the set step distance 1, and the next value of the set value k is obtained as 5.

In the case where the set number k is 5, according to step S4200, as shown in fig. 5, the next similar object list identified by the identification model M1 is obtained to include the object having the highest 5 similarity index values, that is, object 01, object 08, object 05, object 04, and object 02; obtaining a next similar object list identified by the identification model M2 includes 5 objects having highest similarity index values, that is, object 01, object 05, object 02, object 08, and object 04; and obtaining the next similar object list identified by the identification model M3 includes 5 the object having the highest similarity index value, that is, includes object 01, object 08, object 06, object 04, and object 05.

Continuing according to step S4300, a new object list that may be obtained for the corresponding recognition model M1 includes object 02, a new object list that is obtained for the corresponding recognition model M2 includes object 04, and a new object list that is obtained for the corresponding recognition model M3 includes object 05.

After each newly added object list is obtained, reference objects with the current frequency greater than or equal to a set second frequency threshold (2 in this example) can be screened out from all the current similar object lists, and the reference objects include object 01, object 08, object 05 and object 04.

After the reference object is obtained, the judgment is performed according to step S4400 and step S4500, in this example, since the reference object 05 and the reference object 04 appear 1 time in each newly added object list, and the requirement of the first frequency threshold is met, the current value of the set value needs to be updated to be 5 according to step S4500, so that the third recognition operation is performed.

In the third recognition operation, the current value of the value is set to be 5, and each current similar object list obtained in step S4100 is the next similar object list in the second recognition operation, which is not described herein.

Then, the set value k is increased by the set step distance 1, and the next value of the set value k is obtained as 6.

In the case where the set number k is 5, according to step S4200, as shown in fig. 5, the next similar object list identified by the identification model M1 is obtained to include 6 objects having the highest similarity index values, that is, object 01, object 08, object 05, object 04, object 02, and object 18; obtaining a next similar object list identified by the identification model M2 includes 6 objects having highest similarity index values, that is, object 01, object 05, object 02, object 08, object 04, and object 18; and obtaining the next similar object list identified by the identification model M3 includes 5 the object having the highest similarity index value, that is, includes object 01, object 08, object 06, object 04, object 05, and object 14.

Continuing according to step S4300, a new object list for corresponding recognition model M1 may be obtained including object 18, a new object list for corresponding recognition model M2 may be obtained including object 18, and a new object list for corresponding recognition model M3 may be obtained including object 16.

After each newly added object list is obtained, reference objects including object 01, object 08, object 05, object 04, and object 02 can be obtained.

After the reference object is obtained, judgment is made according to step S4400 and step S4500, in which case any reference object does not appear in the newly added object list, and therefore, the final value of the set number k can be determined to be the current value 5.

In the case where the final value of the set number k is determined to be 5, the similar object list identified by each identification model in step S3200 is the current similar object list in the third identification operation for determining the set number k. In this example, according to step S3300, a list of similar objects of the target object 15 may be obtained from the list of similar objects identified by the respective identification models, and the list of similar objects of the target object 15 may be ranked from high to low according to the similarity score, including object 01, object 08, object 04, object 05, and object 02.

According to the embodiment, after the similarity index value of each object in the object set for the target object is obtained through each recognition model, the set value can be adaptively adjusted by the method of the embodiment, so that the set value is the most suitable value for the determined target object and object set and the determined recognition model, in this way, when the similarity index value obtained according to each recognition model is according to the adjusted set number k, the similarity object list obtained by recognition of the corresponding recognition model is screened out from the object set, and when the similarity object list of the target object is obtained according to the similarity object lists, the object dissimilar to the target object is not contained in the finally obtained similarity object list, and the object similar to the target object is not missed, so that the influence of the setting of each value and the threshold on the accuracy of the recognition result is reduced to the greatest extent, and the accuracy of the similarity recognition is effectively improved.

In one embodiment, the method may further comprise: after the target object is acquired according to step S3100, at least two recognition models matching the object type may be selected from the recognition model sets according to the object type of the target object for implementing step S3200 above.

In this embodiment, a set of recognition models may be set, where the set of recognition models includes multiple recognition models, and a mapping table for recording the recognition models that are respectively matched with different object types is stored in advance, so that after determining the target object, a recognition model suitable for performing similar recognition on the target object may be selected according to the object type of the target object and the mapping table.

The object type may be, for example, an item image, a face image, text content, etc.

In this embodiment, since different recognition models have recognition objects with respective good properties, the embodiment selects a plurality of recognition models matching the object type of the target object for similar recognition, which is advantageous for improving the accuracy of the recognition result.

In one embodiment, the method may further comprise: after obtaining the similar object list of the target object according to step S3300, the similar object list of the target object may be output.

The outputting the list of similar objects for the target object includes at least one of displaying the list of similar objects, printing the list of similar objects, and transmitting the list of similar objects to the target device.

For example, the output target object's list of similar objects may include: and sending the similar object list of the target object to terminal equipment bound with a target account, wherein the target account is an account for providing distribution service for the target object and the objects in the object set.

The account number for providing the distribution service is, for example, a work account number of a distributor of the 1# distribution box in the column shown in fig. 1a, or the like.

< method example two >

In this embodiment, a similar identification method is also provided, which can be applied to the logistics distribution scenario as shown in fig. 1 a.

As shown in fig. 6, in this embodiment, the similar identification method may include the following steps S6100 to S6300:

in step S6100, a commodity collection of commodities having at least one same distribution characteristic is obtained.

In step S6100, for example, a commodity set of commodities corresponding to the same delivery box and the same delivery time may be acquired, that is, the acquired commodity set is simultaneously distributed to the same delivery box to perform uniform delivery.

Step S6200, for each commodity in the commodity set, performing similar recognition on the corresponding commodity in the commodity set through each recognition model in at least two recognition models to obtain a list of similar commodities identified by the corresponding commodity.

In this embodiment, it should be understood that, when the similar identification is performed for the corresponding commodity in the commodity set, the similar identification of the corresponding commodity and the same commodity in the commodity set is not required, that is, the other commodities in the commodity set except for the corresponding commodity form the commodity set to be identified of the corresponding commodity.

Step S6300, obtaining a similar commodity list of the corresponding commodity according to the respectively identified similar commodity list.

Therefore, a distributor of the distribution box at the distribution time can know the similar relation of all the commodities in the distribution box according to the similar commodity list, and then the similar commodities are emphasized in distribution, so that accurate distribution is realized.

In one embodiment, the method may further comprise: and sending the obtained similar commodity list to a terminal device bound with a target account, wherein the target account is a distributor account for providing distribution service for the corresponding distribution box at the corresponding distribution time.

In other embodiments, the list of similar products may be printed directly and affixed to the dispensing box for use by the dispenser in identifying similar products during dispensing.

In this logistics distribution scenario, in one embodiment, the similarity identification method may also include the following steps:

step S6100', a commodity collection of commodities having at least one same distribution characteristic is obtained.

Step S6200', for each commodity in the commodity set, obtaining a similar list of the corresponding commodity in the commodity set according to a similar commodity list of the corresponding commodity in the commodity library, which is obtained in advance.

In the step, the similar list of the corresponding commodity in the commodity set can be determined by searching whether other commodities in the commodity set exist in the similar commodity list of the corresponding commodity in the commodity library.

In this embodiment, the obtaining in advance the list of similar commodities in the commodity library corresponding to the commodity may include: respectively carrying out similar recognition on corresponding commodities in the commodity library through each recognition model in at least two recognition models to obtain a similar commodity list which is respectively recognized; and obtaining a similar commodity list of the corresponding commodity in the commodity library according to the respectively identified similar commodity list.

In step S6300', the obtained list of similar lists is provided by the terminal device bound to the target account number, wherein the target account number is a distributor account number providing distribution services for the set of goods.

In this embodiment, the obtained similar commodity list of each commodity in the commodity library in advance may be stored in the server, and when the distributor needs to obtain the similar commodity list of the distributed commodity in the commodity set formed by the distribution box, the distributor may send the commodity list of the commodity set to the server to implement the above steps S6100 'to S6300'. In this regard, in this step S6300', providing the obtained list of similar listings by the terminal device bound to the target account may include: and sending the obtained similar commodity list to terminal equipment bound with the target account.

In this embodiment, the above steps S6100 'to S6300' may also be performed by a terminal device, for example, a terminal device that binds with a target account number. In this regard, the terminal device may obtain, from the server, a similar commodity list of each commodity in the commodity library, which is obtained in advance, to determine a similar commodity list of each commodity in the commodity set, which is not limited herein.

< method example three >

In this embodiment, a similar recognition method is also provided, and this embodiment may be applied to the merchandise push scenario shown in fig. 1 b.

As shown in fig. 7, in this embodiment, the similarity identifying method may include the following steps S7100 to S7400:

step S7100, a commodity selected in the current commodity list is acquired as a target commodity.

The current item list may be a list of items searched for according to a search condition input by a user, the item list including list items of the searched items according to the search condition.

Step S7200, respectively carrying out similar recognition on the target commodity in the current commodity list through each recognition model in at least two recognition models to obtain a similar commodity list respectively recognized.

In this embodiment, it should be understood that, when the similar identification is performed for the corresponding commodity in the commodity list, the similar identification of the corresponding commodity and the same commodity in the commodity list is not required, that is, the other commodities in the commodity list except for the corresponding commodity constitute the commodity set to be identified of the corresponding commodity.

Step S7300, obtaining a list of similar commodities of the target commodity according to the list of similar commodities identified respectively.

Step S7400, displaying the list of similar commodities in the current commodity list.

The step S7400 may be to add a similar goods area displaying similar goods to the current goods list, and display the list item of the goods in the similar goods list in the similar goods area, as shown in fig. 1 b.

In the commodity pushing scenario, in one embodiment, the similarity identifying method may also include the following steps:

step S7100' acquires a commodity selected in the current commodity list as a target commodity.

Step S7200', according to the pre-obtained similar commodity list of the target commodity in the commodity library, obtaining a similar list of the target commodity in the current commodity list.

In the step, the similar list of the corresponding commodity in the current commodity list can be determined by searching whether other commodities in the current commodity list exist in the similar commodity list of the target commodity in the commodity library.

In this embodiment, obtaining in advance a list of similar articles of the target article in the article library may include: respectively carrying out similar recognition on the target commodity in a commodity library through each recognition model in at least two recognition models to obtain a similar commodity list which is respectively recognized; and obtaining a similar commodity list of the target commodity in the commodity library according to the respectively identified similar commodity list.

Step S7300', the list of similar articles is displayed in the current article list.

< device example >

In one embodiment, an apparatus for identifying similar objects is also provided, and FIG. 8 shows a functional block diagram of the apparatus 8000 in one embodiment.

As shown in fig. 8, the apparatus 8000 may include an object acquisition module 6010, an identification processing module 8020, and a manifest generation module 8030.

The object obtaining module 8010 is configured to obtain a target object and an object set to be identified corresponding to the target object.

The recognition processing module 8020 is configured to perform similar recognition on the target object in the object set through each recognition model in the at least two recognition models, so as to obtain a list of similar objects that are recognized respectively.

The list generation module 8030 obtains a list of similar objects of the target object from the respective identified similar object lists.

In one embodiment, the recognition processing module 8020 may be configured to, when performing, by each of at least two recognition models, similar recognition on the target object in the object set, to obtain a list of similar objects that are recognized by each recognition model, respectively: obtaining a similarity index value between each object in the object set and the target object through each identification model; and for each recognition model, obtaining a similar object list recognized by the corresponding recognition model according to the similar index value obtained by the corresponding recognition model.

In one embodiment, the recognition processing module 8020, when obtaining the list of similar objects recognized by the corresponding recognition model according to the similarity index values obtained by the corresponding recognition model, may be configured to: and selecting a set number k objects with the highest similarity index value from the object set according to the similarity index value obtained by the corresponding recognition model, and forming a similar object list recognized by the corresponding recognition model.

In one embodiment, the apparatus 8000 may further include a parameter determination module that, when determining the set number k, may be configured to: obtaining a current similar object list respectively identified by each identification model according to the current numerical value of the set number k; increasing the value of the set number k according to the set step distance to be used as the next value of the set number k; acquiring a next similar object list respectively identified by each identification model when the set number k is the next numerical value; for each recognition model, comparing the corresponding next similar object list with the corresponding current similar object list to obtain a corresponding newly added object list; selecting a reference object from all the current similar object lists; determining that the current value is the final value of the set number k under the condition that the occurrence frequency of each reference object in all the newly added object lists is smaller than the set first frequency threshold value; and updating the current value of the set value to be equal to the next value under the condition that the occurrence frequency of any reference object in all the newly added object lists is greater than or equal to a first frequency threshold value.

In one embodiment, the parameter determination module, when selecting a reference object from all the current similar object lists, may be configured to: and selecting objects with occurrence frequency greater than or equal to a set second frequency threshold value from all the current similar object lists, and referring to the objects as the most.

In one embodiment, the object obtaining module 8010 may be configured to, when obtaining a similarity index value between each object in the object set and the target object: extracting a vector value of a characteristic vector set by a target object for a corresponding recognition model through the corresponding recognition model, and extracting a vector value of each object in the object set for the characteristic vector; and for each object in the object set, obtaining a distance value between the vector value of the corresponding object and the vector value of the target object through the corresponding identification model as a similarity index value between the corresponding object and the target object.

In one embodiment, the manifest generation module 8030, when obtaining a list of similar objects of the target object from the respective identified list of similar objects, may be configured to: for each object in all similar object lists, obtaining a similarity score of the corresponding object according to the occurrence frequency of the corresponding object in all similar object lists; and screening the objects similar to the target object from all the similar object lists according to the similarity scores of each object in all the similar object lists to form a similar object list of the target object.

In one embodiment, the list generation module 8030 may be configured to, when selecting, from all similar object lists, objects similar to the target object according to the similarity score of each object in all similar object lists: in all similar object lists, objects with the similarity score larger than or equal to the set scoring threshold value are screened as objects similar to the target object.

In one embodiment, the object obtaining module 8010, when obtaining the target object and the set of objects to be identified corresponding to the target object, may be configured to: in response to a selection operation for a plurality of objects, selecting any object from the selected objects as the target object, and selecting other objects to form an object set to be identified corresponding to the target object.

In one embodiment, the object obtaining module 8010, when obtaining the target object and the set of objects to be identified corresponding to the target object, may be configured to: in response to a selection operation for any object in the object list, the selected object is selected as a target object, and other objects in the object list are selected to form an object set corresponding to the target object.

In one embodiment, the similar recognition device 8000 may further include an output processing module for outputting a list of similar objects to the target object provided by the list generation module 8030.

In one embodiment, the output processing module, when outputting the list of similar objects of the target object, may be configured to: and sending the similar object list of the target object to terminal equipment bound with a target account, wherein the target account is an account for providing distribution service for the target object and the objects in the object set.

In one embodiment, the apparatus 8000 may further include a model selection module for selecting at least two recognition models from the set of recognition models that match the object type based on the object type of the target object to provide to the recognition processing module 8020.

< device example >

In one embodiment, there is also provided an electronic device that may include an apparatus 8000 for identifying similar objects according to any embodiment of the present invention.

In another embodiment, the electronic device may further comprise a memory and a processor, the memory storing a computer program that, when executed by the processor, implements a similar identification method according to any embodiment of the invention.

In this embodiment, the respective modules of the apparatus 8000 for identifying similar objects described above may be implemented by a processor of an electronic device.

In this embodiment, the electronic device may be, for example, the electronic device 1000 shown in fig. 2. The electronic device may be any terminal device or any server, and is not limited herein.

Embodiments of the present invention also provide a computer readable medium having stored thereon a computer program which, when executed by a processor, implements a similarity identification method according to any of the embodiments of the present invention.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the apparatus, device embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, as relevant to see the section description of the method embodiments.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

The present invention may be a system, method, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for causing a processor to implement aspects of the present invention.

The computer readable storage medium may be a tangible device that can hold and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: portable computer disks, hard disks, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static Random Access Memory (SRAM), portable compact disk read-only memory (CD-ROM), digital Versatile Disks (DVD), memory sticks, floppy disks, mechanical coding devices, punch cards or in-groove structures such as punch cards or grooves having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media, as used herein, are not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., optical pulses through fiber optic cables), or electrical signals transmitted through wires.

The computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device or to an external computer or external storage device over a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers and/or edge servers. The network interface card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium in the respective computing/processing device.

Computer program instructions for carrying out operations of the present invention may be assembly instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, c++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present invention are implemented by personalizing electronic circuitry, such as programmable logic circuitry, field Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), with state information for computer readable program instructions, which can execute the computer readable program instructions.

Various aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable medium having the instructions stored therein includes an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions. It is well known to those skilled in the art that implementation by hardware, implementation by software, and implementation by a combination of software and hardware are all equivalent.

The foregoing description of embodiments of the invention has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the technical improvement of the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.

Claims

1. A method of identifying similar items, comprising:

acquiring a target commodity and a commodity set to be identified corresponding to the target commodity;

obtaining a similarity index value between each commodity in the commodity set and the target commodity through each identification model in at least two identification models;

selecting k commodities with the set number of k which is the highest in the similarity index value from the commodity set according to the similarity index value obtained by the corresponding identification model, and forming a similar commodity list identified by the corresponding identification model;

the set number k is a current value when the occurrence frequency of at least part of commodities in the current similar commodity list in all the newly-added commodity lists is smaller than a set first frequency threshold value; the current similar commodity list is obtained by respectively identifying each identification model according to the current value of the set number k, the new commodity list is obtained by comparing a corresponding next similar commodity list with a corresponding current similar commodity list for each identification model, and the next similar commodity list is obtained by respectively identifying each identification model when the set number k is the next value larger than the current value.

2. The method of claim 1, wherein the method further comprises the step of determining the set number k, comprising:

obtaining a current similar commodity list respectively identified by each identification model according to the current value of the set number k;

acquiring a next similar commodity list respectively identified by each identification model when the set number k is a next numerical value larger than the current numerical value;

Comparing the corresponding next similar commodity list with the corresponding current similar commodity list for each identification model to obtain a corresponding newly-added commodity list;

determining that the current value is the final value of the set number k under the condition that the occurrence frequency of at least part of commodities in the current similar commodity list in all the newly-added commodity lists is smaller than a set first frequency threshold value;

and adjusting the current value of the set value under the condition that the occurrence frequency of any commodity in the at least partial commodities in all the newly-added commodity lists is greater than or equal to the first frequency threshold value.

3. The method of claim 2, wherein the step of determining the set number k further comprises the step of selecting the at least partial commodity, further comprising:

and selecting commodities with occurrence frequency greater than or equal to a set second frequency threshold value from all the current similar commodity lists, and taking the commodities as at least part of the commodities.

4. The method of claim 2, wherein the step of determining the set number k further comprises:

5. The method of claim 2, wherein the adjusting the current value of the set value comprises:

updating the current value of the set value to be equal to the next value.

6. The method of claim 1, wherein the obtaining a similarity index value between each commodity in the set of commodities and the target commodity comprises:

extracting vector values of feature vectors set by the target commodity for the corresponding recognition model through the corresponding recognition model, and extracting vector values of each commodity in the commodity set for the feature vectors;

and for each commodity in the commodity set, obtaining a distance value between the vector value of the corresponding commodity and the vector value of the target commodity through a corresponding identification model, and taking the distance value as a similarity index value between the corresponding commodity and the target commodity.

7. The method of claim 1, wherein the obtaining a list of similar items for the target item from the respective identified list of similar items comprises:

for each commodity in all the similar commodity lists, obtaining a similar score of the corresponding commodity according to the occurrence frequency of the corresponding commodity in all the similar commodity lists;

And screening out commodities similar to the target commodity from all the similar commodity lists according to the similarity scores of all the commodities in the similar commodity lists, so as to form a similar commodity list of the target commodity.

8. The method of claim 7, wherein said screening all of the similar merchandise listings for merchandise similar to the target merchandise based on the similarity score for each of all of the similar merchandise listings comprises:

and screening commodities with the similarity score being greater than or equal to a set score threshold value from all the similar commodity lists as commodities similar to the target commodity.

9. The method of claim 1, wherein the acquiring a target commodity and a set of commodities to be identified corresponding to the target commodity comprises:

and responding to the selection operation of the plurality of commodities, selecting any commodity from the selected plurality of commodities as the target commodity, and selecting other commodities to form a commodity set to be identified corresponding to the target commodity.

10. The method of claim 1, wherein the acquiring a target commodity and a set of commodities to be identified corresponding to the target commodity comprises:

And responding to the selection operation of any commodity in the commodity list, selecting the selected commodity as a target commodity, and selecting other commodities in the commodity list to form a commodity set corresponding to the target commodity.

11. The method of any one of claims 1 to 10, wherein the method further comprises:

and outputting a similar commodity list of the target commodity.

12. The method of claim 11, wherein the outputting the list of similar items of the target item comprises:

and sending the similar commodity list of the target commodity to terminal equipment bound with a target account, wherein the target account is an account for providing distribution service for the target commodity and the commodities in the commodity set.

13. The method of any one of claims 1 to 10, wherein the method further comprises:

and selecting at least two identification models matched with the commodity type from the identification model set according to the commodity type of the target commodity.

14. A method of identifying similar items, comprising:

for each commodity in the commodity set, respectively obtaining a similarity index value between each commodity in the commodity set and a target commodity through each identification model in at least two identification models;

15. The method of claim 14, wherein the acquiring the set of items of merchandise having at least one identical dispensing characteristic comprises:

16. The method of claim 15, wherein the method further comprises:

17. A method of identifying similar items, comprising:

providing an obtained similar list through terminal equipment bound with a target account, wherein the target account is a distributor account for providing distribution service for the commodity set;

obtaining a similarity index value between each commodity in the commodity library and the target commodity through each identification model in at least two identification models;

Selecting k commodities with the set quantity which is the highest in the similarity index value from the commodity library according to the similarity index value obtained by the corresponding identification model, and forming a similar commodity list identified by the corresponding identification model;

obtaining a similar commodity list of the corresponding commodity in the commodity library according to the respectively identified similar commodity list;

18. A method of identifying similar items, comprising:

Respectively obtaining a similarity index value between each commodity in the current commodity list and the target commodity through each identification model in at least two identification models;

selecting k commodities with the set quantity which is the highest in the similarity index value from the current commodity list according to the similarity index value obtained by the corresponding identification model, and forming a similar commodity list identified by the corresponding identification model;

displaying the similar commodity list in the current commodity list;

19. A method of identifying similar items, comprising:

displaying the similar commodity list in the current commodity list;

obtaining a similar commodity list of the target commodity in the commodity library according to the respectively identified similar commodity list;

20. An apparatus for identifying similar objects, comprising:

the commodity acquisition module is used for acquiring target commodities and commodity sets to be identified corresponding to the target commodities;

the identification processing module is used for respectively obtaining the similarity index value between each commodity in the commodity set and the target commodity through each identification model in at least two identification models;

selecting k commodities with the set number of k which is the highest in the similarity index value from the commodity set according to the similarity index value obtained by the corresponding identification model, and forming a similar commodity list identified by the corresponding identification model; the method comprises the steps of,

the list generation module is used for obtaining a similar commodity list of the target commodity according to the similar commodity list identified by each module;

21. An electronic device comprising the apparatus of claim 20; or,

the electronic device includes a memory and a processor;

the memory stores a computer program which, when executed by the processor, performs the method according to any one of claims 1-19.

22. A computer readable storage medium, wherein the computer readable storage medium stores a computer program readable for execution by a computer for performing the method according to any one of claims 1-19 when the computer program is read for execution by the computer.