WO2020119419A1

WO2020119419A1 - Image recognition-based testing and apparatus, and computer device and storage medium

Info

Publication number: WO2020119419A1
Application number: PCT/CN2019/120108
Authority: WO
Inventors: 谭莉
Original assignee: 深圳壹账通智能科技有限公司
Priority date: 2018-12-14
Filing date: 2019-11-22
Publication date: 2020-06-18
Also published as: CN109766915A

Abstract

An image recognition-based testing and apparatus, and a computer device and a storage medium. The method comprises: obtaining a test sample and a recognition object in the test sample according to the unique identifier of a recognition target (s10); receiving a test instruction, obtaining a preset interface address associated with the unique identifier of the recognition target, invoking an image recognition model by means of a preset interface corresponding to the interface address, and inputting the test sample into the image recognition model (s20); obtaining a test sample recognition result output by the image recognition model (s30); and obtaining a test result according to the rate of matching between the recognition result and the recognition object in the test sample (s40). The method increases test efficiency and accuracy, and provides a reference for further adjustment of an image recognition model, thereby improving user experience.

Description

Test method, device, computer equipment and storage medium based on image recognition

This application is based on the Chinese patent application filed on December 14, 2018 with the application number 201811536981.6, titled "Testing Methods, Devices, Computer Equipment, and Storage Media Based on Image Recognition", and claims its priority.

Technical field

The present application relates to the testing field, and in particular to a testing method, device, computer equipment, and storage medium based on image recognition.

Background technique

At present, AI (Artificial Intelligence) technology has gradually emerged and become popular on the Internet. For example, AI technology can be used for image recognition, and image recognition requires learning a large number of images belonging to the same target, so as to train an image recognition model of the target (such as recognition of face images, plant images, etc.), However, in the prior art, after the image recognition model is trained, there are no good test methods and tools to test the image recognition model to check whether the image recognition model meets the requirements of recognizing the target object to determine Whether the image recognition model meets the standards of the current use scene. Therefore, there is an urgent need for a good test method to determine whether the current algorithm in the trained image recognition model can identify the target in a picture or can prompt the user when the target is not included in the picture .

Summary of the invention

The embodiments of the present application provide a test method, device, computer equipment, and storage medium based on image recognition. The present application improves test efficiency and accuracy, and at the same time provides a reference for further adjustment of the image recognition model, improving the user experience.

A test method based on image recognition, including:

Obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

Receiving a test instruction, obtaining a preset interface address associated with the unique identification of the identification target; and invoking an image recognition model through a preset interface corresponding to the interface address, and inputting the test sample into the image recognition model in;

Obtaining the recognition result of the test sample output by the image recognition model;

The test result is obtained according to the matching rate between the recognition result and the recognition object in the test sample.

A test device based on image recognition, including:

The first obtaining module is used to obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

The input module is used to receive a test instruction, obtain a preset interface address associated with the unique identification of the identification target; and call an image recognition model through a preset interface corresponding to the interface address, and input the test sample In the image recognition model;

A second obtaining module, configured to obtain the recognition result of the test sample output by the image recognition model;

The third obtaining module is configured to obtain a test result according to a matching rate between the recognition result and the recognition object in the test sample.

A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the following steps are implemented:

One or more readable storage media storing computer-readable instructions, which when executed by one or more processors, cause the one or more processors to perform the following steps:

The details of one or more embodiments of the present application are set forth in the following drawings and description, and other features and advantages of the present application will become apparent from the description, drawings, and claims.

BRIEF DESCRIPTION

In order to more clearly explain the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application For those of ordinary skill in the art, without paying any creative work, other drawings can also be obtained based on these drawings.

1 is a schematic diagram of an application environment of a test method based on image recognition in an embodiment of the present application;

2 is a flowchart of a test method based on image recognition in an embodiment of the present application;

3 is a flowchart of a test method based on image recognition in another embodiment of the present application;

4 is a flowchart of step S30 of an image recognition-based testing method in an embodiment of the present application;

5 is a flowchart of step S40 of an image recognition-based testing method in an embodiment of the present application;

6 is a flowchart of step S40 of a test method based on image recognition in another embodiment of the present application;

7 is a functional block diagram of a test device based on image recognition in an embodiment of the present application;

8 is a schematic block diagram of a test device based on image recognition in another embodiment of the present application;

9 is a schematic block diagram of a second acquisition module of a test device based on image recognition in an embodiment of the present application;

10 is a schematic diagram of a computer device in an embodiment of the present application.

detailed description

The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

The test method based on image recognition provided by this application calls the image recognition model to be tested through a preset interface, so as to perform a full-scale test on the called image recognition model with a test sample to verify whether the image recognition model is achieved through the test results It can identify the requirements of the target, and then determine whether the image recognition model meets the standard of the current use scene; when the image recognition model does not meet the standard of the current use scene, the image recognition model can also be based on the test result Make adjustments. This application improves the testing efficiency and accuracy, and at the same time provides a reference for the further adjustment of the image recognition model and improves the user experience. This application can be applied in an application environment as shown in FIG. 1, in which a client (computer device) communicates with a server through a network. Among them, the client (computer device) includes but is not limited to various personal computers, notebook computers, smart phones, tablet computers, cameras and portable wearable devices. The server can be realized by an independent server or a server cluster composed of multiple servers.

In an embodiment, as shown in FIG. 2, a test method based on image recognition is provided, and the method is applied to the server in FIG. 1 as an example for illustration, including the following steps:

S10. Acquire a test sample and the identification object in the test sample according to the unique identification of the identification target.

The test sample refers to a sample used to test whether the image recognition model can accurately recognize the recognition target.

The recognition target refers to a recognition target of the image recognition model. For example, if the image recognition model is used to recognize a peony flower, the recognition target is a peony flower. The current test samples are all for testing the image Whether the identification model can identify peony flowers.

The unique identification of the identification target refers to a unique identification such as the number and name of the identification target.

The recognition object refers to an object included in the test sample that can be recognized by the image recognition model; each of the test samples includes at least one test object, the recognition object may be the same as the recognition target, or may be The recognition target is different. For example, the recognition target is peony, but the type of a test sample is an interference sample. At this time, the test object in the interference sample may be peony flowers similar to peony flowers.

In an embodiment, as shown in FIG. 3, before the step S10, the following steps are included:

S101: Create a test sample library, set sample parameters in the test sample library, and store the sample test library in association with a unique identifier that identifies a target;

The sample parameters include, but are not limited to: the number and type of test samples, the scene in which the identification target included in the test sample is located, the angle of the identification target displayed in the test sample, and the like.

Wherein, the test sample contains a recognition object (that is, a recognition object that can be recognized by the image recognition model), and the type of the test sample includes a conforming sample (the test sample contains a recognition object consistent with the recognition target, For example, if the identification target is peony, the identification object included in the test sample is also peony), and the non-conforming sample (the test sample contains identification objects that are inconsistent with the identification target, for example, if the identification target is peony, the test sample The identification object included in the leaf is a leaf), interference samples (the test sample contains identification objects that are inconsistent with the identification target, and are similar to the identification target and are easily confused, for example, the identification target is peony, the test sample contains The object of identification is peony similar to the peony flower).

The scene in which the recognition target included in the test sample refers to different recognition scenes in which the recognition target must be included in the conforming sample, for example, the test sample includes the recognition target in day and night , Solid color background, complex background and other test samples in different recognition scenarios. Understandably, in some embodiments, the scene in which the identified object in the non-compliant sample and the interference sample may be set.

The angle of the identification target displayed in the test sample refers to the display angle of the identification target that must be included in the conforming sample, for example, the display angle of the identification target in the test sample includes upward and downward incline ,, left and right, and you can set the specific angle range of their inclination (you can set different angle ranges for the same angle inclination, for example, you can set the sample parameters to include uptilt angles of 0-30 degrees, 30- 60 degrees, 60-90 degrees three incline angles). Understandably, in some embodiments, the display angle of the identification object in the non-conforming sample and the interference sample may also be set.

S102: Receive a test sample uploaded by the user to the test sample library on the client, record the identification object included in the test sample, and detect whether the uploaded test sample matches the sample parameter;

Understandably, the test samples require users (such as users or testers who create image recognition models) to upload them on the client.

After uploading the test samples, it will automatically detect whether all uploaded test samples conform to the sample parameters set in step S101 above, for example, whether the types of test samples and the number of various types of test samples meet the requirements. Contains the identification target scene and whether the number of test samples of each scene meets the requirements, whether the angle of the identification target displayed in the test sample and the number of test samples at each angle meet the requirements, etc.; if the uploaded test sample meets all For sample parameters, it can be determined that the uploaded test sample matches the sample parameter, and if the uploaded test sample does not match one of the sample parameters, it is determined that the uploaded test sample does not match the sample parameter.

In an embodiment, a parameter identification model is established for each of the sample parameters to detect whether the uploaded test sample matches the sample parameter selected by the user when uploading; for example, if the user's currently uploaded test sample is the identification included in the test sample The scene where the target is located is a coincident sample during the day. At this time, the parameter identification model needs to identify whether the scene in the test sample is daytime, and whether the identification object in the test sample is the identification target; if it is daytime and the identification object is Identify the target, at this time it is considered that the uploaded test sample matches the sample parameter selected by the user when uploading; if one of the items does not match (the scene is not daytime, or the identification object in the test sample is not the identification target), at this time, the uploaded The test sample does not match the sample parameter selected by the user when uploading, prompting the user to re-upload or suggesting that there is a test risk. Understandably, in another embodiment, the test samples uploaded by the default user are all test samples whose confirmation is consistent with the sample parameters, and are no longer tested to reduce the server load and improve the test efficiency.

S103, when the test sample matches the sample parameter, the test sample is prompted to meet the requirements;

That is, if the uploaded test sample meets all sample parameters (the number and type of each test sample uploaded by the user all meet the requirements), the uploaded test sample matches the sample parameter, and at this time, the user is prompted to upload Of the test samples meet the requirements, and uploading the test samples is successful.

S104: When the test sample does not match the sample parameter, display content that the test sample does not match the sample parameter and/or prompt that there is a test risk.

That is, if the uploaded test sample does not match one or more of the sample parameters, it is determined that the uploaded test sample does not match the sample parameter. At this time, the user is prompted to have a test risk, such as prompting the user To the test effect"; at the same time, it can also display the specific content of the unmatched sample parameters, for example, display "the number of interference samples is still missing 10".

In an embodiment, the step S10 includes the following steps:

The test samples in the test sample library are retrieved according to the unique identification of the identification target, and the identification objects in the test samples are obtained.

That is, if testing is required, you can enter (or select from the list of recognition targets) the unique identifier of the recognition target on the current display interface (for example, if the recognition target is peony, you can select it in the current display interface) The test target is the unique number of the peony flower), and the test sample in the test sample library is retrieved according to the unique identification of the identification target; as a preference, in one embodiment, the front end of the webpage for testing in the client The test sample library can be uploaded on the webpage. In the front-end webpage, when the user confirms that a certain identification target needs to be tested, the user can click the preset upload button on the front-end webpage. At this time, an input will appear on the front-end webpage. The box is for the user to input the link address of the sample library associated with the identification target. After the user fills in the link address, he can jump to the sample library corresponding to the link address to obtain the test sample in the sample library to complete the upload test The sample library process.

Understandably, the test sample library may be stored in the background database in the background server, at this time it is only necessary to directly retrieve the test sample from the background database according to the unique identification of the identification target; in another embodiment The test sample library is stored in the local database of the client that creates the image recognition model. At this time, it is necessary to first obtain the test sample from the local database according to the unique identifier of the recognition target to the background database, so as to facilitate subsequent testing.

Understandably, no matter where the test sample library is, the unique identification of the identification target is associated with the address of the test sample library. After selecting the unique identification of the identification target, the test sample library can be obtained And enter the test sample library corresponding to the address to retrieve test samples.

At the same time, since the test samples all contain the identification objects, when the test samples are acquired, the identification objects in the test samples are acquired at the same time.

S20. Receive a test instruction, obtain a preset interface address associated with the unique identification of the identification target, and call an image recognition model through a preset interface corresponding to the interface address, and input the test sample into the image In the recognition model;

In this embodiment, the server provides a preset interface to facilitate the tester to call the image recognition model to be tested through the interface; wherein, the test instruction refers to the tester inputting or selecting a recognition target associated with the image recognition model on the client Unique identification (the unique identification of the identification target is associated with the interface address of the preset interface of the image identification model, so after entering or selecting the unique identification of the identification target, the image identification model call can be obtained Preset the interface address of the interface, so that after calling the image recognition model through the preset interface), when it is confirmed that a test needs to be performed (the test sample has been obtained at this time), it is sent by clicking or sliding a preset button Test instructions. Understandably, the above interface address can be entered in the front-end webpage. For example, after selecting the unique identification of the identification target, it will automatically jump to the front-end webpage. If the tester has input the To uniquely identify the associated interface address, the interface address is automatically displayed at this time. If the interface address has not been entered before, the interface address can be manually entered or selected and associated with the unique identifier of the identification target.

The server calls the image recognition model through the preset interface to perform image recognition on each of the test samples, and starts the test process.

S30. Obtain the recognition result of the test sample output by the image recognition model.

In this embodiment, the image recognition model inputs the test sample transmitted through a preset interface, and the image recognition model outputs the recognition result of the test sample. The image recognition model There are only two kinds of recognition results: the test sample contains a recognition target (in this case, the image recognition model outputs a pass) and the test sample does not contain a recognition target (in this case, the image recognition model outputs a fail).

In an embodiment, as shown in FIG. 4, in step S30, the acquiring the recognition result of the test sample output by the image recognition model includes:

S301. Extract the first image feature of the identification object of the test sample from the test samples input in the image recognition model; that is, extract the first image of the identification object in each of the test samples feature.

S302: Retrieve the second image feature of the recognition target according to the unique identifier of the recognition target; that is, each image recognition model is associated with a recognition target (or a second image feature extracted from the recognition target ), if only the recognition target is associated, in this step, the second image feature needs to be extracted from the recognition target.

S303. Determine, through the image recognition model, whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold; that is, the image recognition model can obtain the first image feature and The similarity of the second image feature (including the color, shape, contour, distance between each feature and combination method of each feature in the image feature to determine), and the similarity threshold can be set according to demand, For example, set between 0.6-1.

S304: When the similarity between the first image feature and the second image feature exceeds a preset similarity threshold, output a recognition result including the recognition target in the test sample; that is, in the first image feature When the similarity with the second image feature exceeds a preset similarity threshold, the identification object in the test sample and the identification target are considered to be the same object, and at this time, the test object in the test sample is considered to be Identify the target, and output the identification result of the test sample containing the identified target.

S305. When the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold, output a recognition result that the test sample does not include a recognition target. That is, when the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold, it is considered that the recognition object in the test sample and the recognition target are not the same object, at this time, It is considered that the test object in the test sample is not the recognition target, and the recognition result that the test sample does not contain the recognition target is output.

S40. Obtain a test result according to a matching rate between the recognition result and the recognition object in the test sample.

Wherein, the test result refers to a determination result of determining whether the image recognition model meets the standard. The test result may include that the recognition accuracy of the image recognition model has reached or has not yet reached the standard.

In an embodiment, the recognition result includes that the test sample includes a recognition target and that the test sample does not include a recognition target; as shown in FIG. 5, the step S40 includes the following steps:

S401: Detect whether the recognition result matches the recognition object in the test sample.

That is, if the identification object in the test sample previously recorded in step S102 is the same as the identification target associated with the test sample library, it means that the test sample contains the identification target; at this time, if The identification result for the test sample is that the test sample contains the identification target, that is, it is detected that the identification result matches the identification object in the test sample; if the identification result for the test sample is the The identification target is not included in the test sample, that is, it is detected that the identification result does not match the identification object in the test sample.

If the identification object in the test sample previously recorded in step S102, the identification target associated with the test sample library is different (such as an interference sample or a non-conforming sample), it means that the test sample does not contain The identification target; at this time, if the identification result for the test sample is that the test sample contains the identification target, it is detected that the identification result does not match the identification object in the test sample; if The identification result of the test sample is that the test sample does not include the identification target, that is, it is detected that the identification result matches the identification object in the test sample.

S402, when the recognition result matches the recognition object in the test sample, the test sample is counted as a matching sample.

S403, when the recognition result does not match the recognition object in the test sample, the test sample is counted as a mismatch sample.

S404. Calculate the matching rate between the recognition result and the recognition object in the test sample according to the following formula:

X=A/(A+B)*100%

among them:

X is the matching rate between the recognition result and the recognition object in the test sample;

A is the matching sample;

B is the unmatched sample.

S405: Obtain a test result according to the matching rate, where the test result includes the image recognition model test passing and the image recognition model test failing.

When the test result is that the image recognition model has passed the test, the user who created the image recognition model may be prompted, the image recognition model has reached the standard, and the image recognition model may be used for image recognition; when the test result is When the image recognition model test fails, the user is prompted to further adjust the image recognition model.

In an embodiment, as shown in FIG. 6, after the step S405, the method further includes:

S406, when the test result is that the image recognition model test fails, it prompts that the image recognition model needs to be further adjusted.

S407: Obtain the sample parameters corresponding to each of the unmatched samples, count the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, Sort the sample parameters; that is, in this step, obtain the specific data of the unmatched sample and the matched sample in the test sample, and determine the worst training effect in the image recognition model based on the above data And adjust it.

S408: Determine the link with the worst training effect of the image recognition model according to the ranking results of the sample parameters, and adjust the link with the worst training effect of the image recognition model. That is, the greater the number of unmatched samples corresponding to the sample parameters, the worse the training effect of the image recognition model on the test samples of the sample parameters. Conversely, the unmatched corresponding to the sample parameters The smaller the number of samples, the better the training effect of the image recognition model on the test samples of the sample parameters; after determining the preset number of links with the worst training effect of the image recognition model, the training can be strengthened for this link To obtain a more complete image recognition model, and then complete the adjustment of the image recognition model.

In an embodiment, as shown in FIG. 7, a test device based on image recognition is provided. The test device based on image recognition corresponds one-to-one with the test method based on image recognition in the above embodiment. The test device based on image recognition includes:

The first obtaining module 11 is configured to obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

The input module 12 is used to receive a test instruction, obtain a preset interface address associated with the unique identification of the identification target; and call an image recognition model through a preset interface corresponding to the interface address, and transfer the test sample Input into the image recognition model;

The second obtaining module 13 is configured to obtain the recognition result of the test sample output by the image recognition model;

The third obtaining module 14 is configured to obtain a test result according to a matching rate between the recognition result and the recognition object in the test sample.

In an embodiment, as shown in FIG. 8, the device further includes:

The creation module 15 is used to create a test sample library, set the sample parameters in the test sample library, and store the sample test library in association with the unique identification of the identification target;

The detection module 16 is configured to receive a test sample uploaded by the user to the test sample library on the client, record the identification object included in the test sample, and detect whether the uploaded test sample matches the sample parameter;

The prompt module 17 is used to prompt the test sample to meet the requirements when the test sample matches the sample parameter;

The display module 18 is configured to display content that the test sample does not match the sample parameter and/or indicate that there is a test risk when the test sample does not match the sample parameter.

In an embodiment, as shown in FIG. 9, the second obtaining module 13 includes:

The extraction unit 131 is configured to extract the first image feature of the identification object of the test sample from the test samples input in the image recognition model;

The retrieval unit 132 is configured to retrieve the second image feature of the recognition target according to the unique identifier of the recognition target;

The judging unit 133 is configured to judge whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold through the image recognition model;

The first output unit 134 is configured to output a recognition result containing a recognition target in the test sample when the similarity between the first image feature and the second image feature exceeds a preset similarity threshold;

The second output unit 135 is configured to output a recognition result that does not include a recognition target in the test sample when the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold.

In an embodiment, the third obtaining module 14 includes:

A detection unit, configured to detect whether the recognition result matches the recognition object in the test sample;

A first counting unit for counting the test sample into the matching sample when the recognition result matches the identification object in the test sample;

A second counting unit, configured to count the test sample as a non-matching sample when the recognition result does not match the identification object in the test sample;

The calculation unit is used to calculate the matching rate between the recognition result and the recognition object in the test sample according to the following formula:

X=A/(A+B)*100%

among them:

A is the matching sample;

B is the unmatched sample;

The obtaining unit is configured to obtain a test result according to the matching rate, where the test result includes the image recognition model test passing and the image recognition model test failing.

In an embodiment, the third obtaining module 14 further includes:

A prompting unit, for prompting that the image recognition model needs to be further adjusted when the test result is that the image recognition model test fails;

A statistical unit, configured to obtain sample parameters corresponding to each of the unmatched samples, count the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, Sort each of the sample parameters;

The adjusting unit is configured to determine the link with the worst training effect of the image recognition model according to the ranking results of the sample parameters, and adjust the link with the worst training effect of the image recognition model.

For the specific definition of the test device based on image recognition, reference may be made to the above definition of the test method based on image recognition, which will not be repeated here. Each module in the above-mentioned image recognition-based test device may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in the hardware or independent of the processor in the computer device, or may be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure may be as shown in FIG. 10. The computer device includes a processor, memory, network interface, and database connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and internal memory. The readable storage medium stores an operating system, computer readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. . The computer-readable instructions are executed by the processor to implement a test method based on image recognition. The readable storage medium provided by this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor. When the processor executes the computer-readable instructions, the following steps are implemented:

In one embodiment, one or more readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage Medium; the computer-readable instructions are stored on the readable storage medium, and when the computer-readable instructions are executed by one or more processors, the one or more processors implement the following steps:

A person of ordinary skill in the art may understand that all or part of the process in the method of the foregoing embodiments may be completed by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions may be stored in a non-volatile computer In a readable storage medium or a volatile readable storage medium, when the computer-readable instructions are executed, they may include the processes of the foregoing method embodiments. Wherein, any reference to the memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Those skilled in the art can clearly understand that, for convenience and conciseness of description, only the above-mentioned division of each functional unit or module is used as an example for illustration. In practical applications, the above-mentioned functions may be allocated by different functional units or Module completion means that the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still implement the foregoing The technical solutions described in the examples are modified, or some of the technical features are equivalently replaced; and these modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application. Within the scope of protection of this application.

Claims

A test method based on image recognition is characterized by including:

Obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

Receiving a test instruction, obtaining a preset interface address associated with the unique identification of the identification target; and invoking an image recognition model through a preset interface corresponding to the interface address, and inputting the test sample into the image recognition model in;

Obtaining the recognition result of the test sample output by the image recognition model;

The test result is obtained according to the matching rate between the recognition result and the recognition object in the test sample.
The test method based on image recognition according to claim 1, wherein before acquiring the test sample and the identification object in the test sample according to the unique identification of the identification target, comprising:

Create a test sample library, set sample parameters in the test sample library, and store the sample test library in association with a unique identification of the identification target;

Receiving a test sample uploaded by the user to the test sample library on the client, recording the identification object included in the test sample, and detecting whether the uploaded test sample matches the sample parameter;

When the test sample matches the sample parameter, the test sample is prompted to meet the requirements;

When the test sample does not match the sample parameter, display content that the test sample does not match the sample parameter and/or suggest that there is a test risk.
The test method based on image recognition according to claim 1, wherein the obtaining the recognition result of the test sample output by the image recognition model includes:

Extracting the first image feature of the identification object of the test sample from the test sample input in the image recognition model;

Extracting the second image feature of the identification target according to the unique identification of the identification target;

Judging by the image recognition model whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold;

When the similarity between the first image feature and the second image feature exceeds a preset similarity threshold, output a recognition result including a recognition target in the test sample;

When the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold, output a recognition result that does not include a recognition target in the test sample.
The test method based on image recognition according to claim 2, wherein the obtaining the test result according to the matching rate between the recognition result and the recognition object in the test sample includes:

Detecting whether the recognition result matches the recognition object in the test sample;

When the recognition result matches the recognition object in the test sample, the test sample is counted into the matching sample;

When the recognition result does not match the recognition object in the test sample, the test sample is counted as a non-matching sample;

The matching rate between the recognition result and the recognition object in the test sample is calculated according to the following formula:

X=A/(A+B)*100%

among them:

X is the matching rate between the recognition result and the recognition object in the test sample;

A is the matching sample;

B is the unmatched sample;

A test result is obtained according to the matching rate, and the test result includes that the image recognition model test passes and the image recognition model test fails.
The test method based on image recognition according to claim 4, wherein after the test result is obtained according to the matching rate, the method further comprises:

When the test result is that the image recognition model test fails, it prompts that the image recognition model needs to be further adjusted;

Obtaining sample parameters corresponding to each of the unmatched samples, counting the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, for each sample Parameter sorting;

The link with the worst training effect of the image recognition model is determined according to the ranking results of the sample parameters, and the link with the worst training effect of the image recognition model is adjusted.
A test device based on image recognition, characterized in that it includes:

The first obtaining module is used to obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

The input module is used to receive a test instruction, obtain a preset interface address associated with the unique identification of the identification target; and call an image recognition model through a preset interface corresponding to the interface address, and input the test sample In the image recognition model;

A second obtaining module, configured to obtain the recognition result of the test sample output by the image recognition model;

The third obtaining module is configured to obtain a test result according to a matching rate between the recognition result and the recognition object in the test sample.
The test device based on image recognition according to claim 6, wherein the test device further comprises:

A creation module, used to create a test sample library, set the sample parameters in the test sample library, and store the sample test library in association with the unique identification of the identification target;

The detection module is configured to receive the test sample uploaded by the user to the test sample library on the client, record the identification object included in the test sample, and detect whether the uploaded test sample matches the sample parameter;

A prompt module, used to prompt the test sample to meet requirements when the test sample matches the sample parameter;

The display module is configured to display content that the test sample does not match the sample parameter and/or indicate that there is a test risk when the test sample does not match the sample parameter.
The test device based on image recognition according to claim 6, wherein the second acquisition module comprises:

An extraction unit, configured to extract the first image feature of the identification object of the test sample from the test samples input in the image identification model;

A retrieving unit, configured to retrieve the second image feature of the recognition target according to the unique identifier of the recognition target;

A judgment unit, configured to judge whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold through the image recognition model;

A first output unit, configured to output a recognition result including a recognition target in the test sample when the similarity between the first image feature and the second image feature exceeds a preset similarity threshold;

The second output unit is configured to output a recognition result that does not include a recognition target in the test sample when the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold.
The test device based on image recognition according to claim 7, wherein the third acquisition module comprises:

A detection unit, configured to detect whether the recognition result matches the recognition object in the test sample;

A first counting unit for counting the test sample into the matching sample when the recognition result matches the identification object in the test sample;

A second counting unit, configured to count the test sample as a non-matching sample when the recognition result does not match the identification object in the test sample;

The calculation unit is used to calculate the matching rate between the recognition result and the recognition object in the test sample according to the following formula:

X=A/(A+B)*100%

among them:

X is the matching rate between the recognition result and the recognition object in the test sample;

A is the matching sample;

B is the unmatched sample;

The obtaining unit is configured to obtain a test result according to the matching rate, where the test result includes the image recognition model test passing and the image recognition model test failing.
The test device based on image recognition according to claim 9, wherein the third acquisition module further comprises:

A prompting unit, for prompting that the image recognition model needs to be further adjusted when the test result is that the image recognition model test fails;

A statistical unit, configured to obtain sample parameters corresponding to each of the unmatched samples, count the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, Sort each of the sample parameters;

The adjusting unit is configured to determine the link with the worst training effect of the image recognition model according to the ranking results of the sample parameters, and adjust the link with the worst training effect of the image recognition model.
A computer device, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that, when the processor executes the computer-readable instructions, it is implemented as follows step:

Obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

Receiving a test instruction, obtaining a preset interface address associated with the unique identification of the identification target; and invoking an image recognition model through a preset interface corresponding to the interface address, and inputting the test sample into the image recognition model in;

Obtaining the recognition result of the test sample output by the image recognition model;

The test result is obtained according to the matching rate between the recognition result and the recognition object in the test sample.
The computer device according to claim 11, wherein the processor executes the computer-readable instruction before acquiring the test sample and the identification object in the test sample according to the unique identification of the identification target Implement the following steps:

Create a test sample library, set sample parameters in the test sample library, and store the sample test library in association with a unique identification of the identification target;

Receiving a test sample uploaded by the user to the test sample library on the client, recording the identification object included in the test sample, and detecting whether the uploaded test sample matches the sample parameter;

When the test sample matches the sample parameter, the test sample is prompted to meet the requirements;

When the test sample does not match the sample parameter, display content that the test sample does not match the sample parameter and/or suggest that there is a test risk.
The computer device according to claim 11, wherein the obtaining the recognition result of the test sample output by the image recognition model comprises:

Extracting the first image feature of the identification object of the test sample from the test sample input in the image recognition model;

Extracting the second image feature of the identification target according to the unique identification of the identification target;

Judging by the image recognition model whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold;

When the similarity between the first image feature and the second image feature exceeds a preset similarity threshold, output a recognition result including the recognition target in the test sample;

When the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold, output a recognition result that does not include a recognition target in the test sample.
The computer device according to claim 12, wherein the obtaining the test result according to the matching rate between the recognition result and the recognition object in the test sample includes:

Detecting whether the recognition result matches the recognition object in the test sample;

When the recognition result matches the recognition object in the test sample, the test sample is counted into the matching sample;

When the recognition result does not match the recognition object in the test sample, the test sample is counted as a non-matching sample;

The matching rate between the recognition result and the recognition object in the test sample is calculated according to the following formula:

X=A/(A+B)*100%

among them:

X is the matching rate between the recognition result and the recognition object in the test sample;

A is the matching sample;

B is the unmatched sample;

A test result is obtained according to the matching rate, and the test result includes that the image recognition model test passes and the image recognition model test fails.
The computer device according to claim 14, wherein after the test result is obtained according to the matching rate, when the processor executes the computer-readable instruction, the following steps are further implemented:

When the test result is that the image recognition model test fails, it prompts that the image recognition model needs to be further adjusted;

Obtaining sample parameters corresponding to each of the unmatched samples, counting the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, for each sample Parameter sorting;

The link with the worst training effect of the image recognition model is determined according to the ranking results of the sample parameters, and the link with the worst training effect of the image recognition model is adjusted.
One or more readable storage media storing computer readable instructions, characterized in that, when the computer readable instructions are executed by one or more processors, the one or more processors execute the following steps:

Obtain the test sample and the identification object in the test sample according to the unique identification of the identification target;

Receiving a test instruction, obtaining a preset interface address associated with the unique identification of the identification target; and invoking an image recognition model through a preset interface corresponding to the interface address, and inputting the test sample into the image recognition model in;

Obtaining the recognition result of the test sample output by the image recognition model;

The test result is obtained according to the matching rate between the recognition result and the recognition object in the test sample.
The readable storage medium according to claim 16, wherein the computer-readable instruction is one or more before acquiring the test sample and the identification object in the test sample according to the unique identification of the identification target When the processor executes, the one or more processors further execute the following steps:

Create a test sample library, set sample parameters in the test sample library, and store the sample test library in association with a unique identification of the identification target;

Receiving a test sample uploaded by the user to the test sample library on the client, recording the identification object included in the test sample, and detecting whether the uploaded test sample matches the sample parameter;

When the test sample matches the sample parameter, the test sample is prompted to meet the requirements;

When the test sample does not match the sample parameter, display content that the test sample does not match the sample parameter and/or suggest that there is a test risk.
The readable storage medium according to claim 16, wherein the acquiring the recognition result of the test sample output by the image recognition model includes:

Extracting the first image feature of the identification object of the test sample from the test sample input in the image recognition model;

Retrieving the second image feature of the identification target according to the unique identification of the identification target;

Judging by the image recognition model whether the similarity between the first image feature and the second image feature exceeds a preset similarity threshold;

When the similarity between the first image feature and the second image feature exceeds a preset similarity threshold, output a recognition result including a recognition target in the test sample;

When the similarity between the first image feature and the second image feature does not exceed a preset similarity threshold, output a recognition result that does not include a recognition target in the test sample.
The readable storage medium according to claim 17, wherein the obtaining the test result according to the matching rate between the recognition result and the recognition object in the test sample includes:

Detecting whether the recognition result matches the recognition object in the test sample;

When the recognition result matches the recognition object in the test sample, the test sample is counted into the matching sample;

When the recognition result does not match the recognition object in the test sample, the test sample is counted as a non-matching sample;

The matching rate between the recognition result and the recognition object in the test sample is calculated according to the following formula:

X=A/(A+B)*100%

among them:

X is the matching rate between the recognition result and the recognition object in the test sample;

A is the matching sample;

B is the unmatched sample;

A test result is obtained according to the matching rate, and the test result includes that the image recognition model test passes and the image recognition model test fails.
The readable storage medium of claim 19, wherein after the test result is obtained according to the matching rate, when the computer-readable instructions are executed by one or more processors, the one or more Each processor also performs the following steps:

When the test result is that the image recognition model test fails, it prompts that the image recognition model needs to be further adjusted;

Obtaining sample parameters corresponding to each of the unmatched samples, counting the number of the unmatched samples corresponding to each of the sample parameters, and according to the number of the unmatched samples corresponding to each of the sample parameters, for each sample Parameter sorting;

The link with the worst training effect of the image recognition model is determined according to the ranking results of the sample parameters, and the link with the worst training effect of the image recognition model is adjusted.