Model evaluation method, device and computer equipment

Abstract

The present invention discloses a model evaluation method, device and computer equipment for evaluating the performance of a picture comparison model; the method comprises: obtaining a sample set, the sample set comprising output data of the picture comparison model; the output data of the picture comparison model at least comprising first picture data, second picture data associated with the first picture data and a matching value between the first picture data and the second picture data; calculating the comparison accuracy of the picture comparison model and the picture recall rate of the picture comparison model based on the sample set; evaluating the performance of the picture comparison model based on the comparison accuracy and the picture recall rate; sampling and detecting the comparison results of the picture comparison model, and evaluating the comparison accuracy and the picture recall rate indicators of the picture comparison model according to the sampled sample set, thereby quickly, stably and accurately evaluating the performance of the picture comparison model and reducing the workload of manual judgment and calculation.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a model evaluation method, device and computer equipment. Background Art

[0002] E-commerce platforms sell a variety of products, and product details pages usually display product images for consumers to browse. In order to control the quality of the products sold, e-commerce platforms need to compare and detect the images of the massive products they sell with those sold on rival platforms. This comparison and detection task is performed by the image comparison model.

[0003] In order to ensure the reliability of the comparison and detection tasks, it is necessary to ensure that the performance of the image comparison model performing the comparison and detection tasks is excellent, so the performance of the image comparison model needs to be evaluated frequently. The current solution for the performance evaluation of the image comparison model is to manually extract the comparison and detection result samples of some image comparison models, and then spend a lot of manpower on manual judgment and calculation after sampling. The evaluation efficiency is low, the number of samples extracted by manual sampling is limited, and the reliability of the evaluation results is low.

[0004] Therefore, an evaluation method is needed that can sample data online and run data online independently without relying on an offline environment, thereby avoiding the problem of small sampling quantities and reducing the workload of manual judgment and calculation. Summary of the invention

[0005] The purpose of the present invention is to provide a model evaluation method, device and computer equipment that can sample data online and calculate evaluation indicators online without relying on an offline environment, thereby avoiding the problem of small sampling quantity and reducing the workload of manual judgment and calculation.

[0006] The technical solution of the present invention is: in a first aspect, the present invention provides a model evaluation method for evaluating the performance of an image comparison model; the method comprises:

[0007] Acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data;

[0008] Calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set;

[0009] The image comparison model performance is evaluated based on the comparison accuracy, the image recall rate and preset evaluation rules.

[0010] In a preferred embodiment, the obtaining of the sample set includes:

[0011] Obtaining output data of the image comparison model within a preset time period;

[0012] The output data of the image comparison model within the preset time period is sampled to obtain a sample set.

[0013] In a preferred embodiment, the step of sampling the output data of the image comparison model within the preset time period to obtain a sample set includes:

[0014] Sampling the output data of the image comparison model within the preset time period to obtain sampled data;

[0015] The sampled data is organized into an EXCEL file, and the sample set includes the EXCEL file.

[0016] In a preferred embodiment, the sampling of the output data of the image comparison model within the preset time period to obtain the sampled data includes:

[0017] The output data of the image comparison model within the preset time period is sampled by commodity category to obtain sampling data of each commodity category.

[0018] In a preferred embodiment, before calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set, the method further includes:

[0019] Filtering the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the first preset condition to obtain a first sub-sample set;

[0020] The output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the second preset condition is screened to obtain a second sub-sample set.

[0021] In a preferred embodiment, the calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set includes:

[0022] Querying a database to obtain a barcode corresponding to each piece of first image data and second image data in the first sub-sample set;

[0023] Obtaining a comparison accuracy rate of the image comparison model based on the barcodes corresponding to the first image data and the second image data;

[0024] After a second preset time period, the second sub-sample set is sent to the image comparison model to obtain the number of recalled images, where the recalled images are the number of images whose matching values ​​with the first image data reach a preset threshold, obtained by the image comparison model from the current dynamic product image pool according to the first image data in the second sub-sample set;

[0025] The image recall rate of the image comparison model is obtained based on the number of images whose matching values ​​with the first image data reach a preset threshold.

[0026] In a preferred embodiment, the obtaining the comparison accuracy of the image comparison model based on the barcodes corresponding to the first image data and the second image data includes:

[0027] Performing domain determination based on the barcodes corresponding to the first image data and the second image data;

[0028] If the barcodes corresponding to the first image data and the second image data are the same, then the image comparison model matches correctly;

[0029] Counting the number of correct matches of the image comparison model;

[0030] The comparison accuracy of the picture comparison model is obtained by calculating the ratio of the number of correct matches of the picture comparison model to the output data volume of the picture comparison model in the first sub-sample set.

[0031] In a preferred embodiment, obtaining the image recall rate of the image comparison model based on the number of images whose matching values ​​with the first image data reach a preset threshold comprises:

[0032] The image recall rate of the image comparison model is obtained by calculating the ratio of the number of images whose matching value with the first image data is greater than or equal to a preset threshold and the number of the second image data in the second sub-sample set.

[0033] In a second aspect, the present invention provides a model evaluation device for evaluating the performance of an image comparison model, the device comprising:

[0034] an acquisition module, configured to acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data;

[0035] A calculation module, used for calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set;

[0036] An evaluation module is used to evaluate the performance of the image comparison model based on the comparison accuracy, the image recall rate and preset evaluation rules.

[0037] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0038] Compared with the prior art, the advantages of the present invention are: providing a model evaluation method, device and computer equipment for evaluating the performance of an image comparison model; the method comprises: obtaining a sample set, the sample set comprising output data of the image comparison model; the output data of the image comparison model comprising at least first image data, second image data associated with the first image data and a matching value between the first image data and the second image data; calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set; evaluating the performance of the image comparison model based on the comparison accuracy, the image recall rate and preset evaluation rules; sampling and detecting the comparison results of the image comparison model, and evaluating the comparison accuracy and image recall rate indicators of the image comparison model according to the sampled sample set, thereby quickly, stably and accurately evaluating the performance of the image comparison model and reducing the workload of manual judgment and calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0041] Figure 1 A flow chart of a model evaluation method provided by an embodiment of the present invention;

[0042] Figure 2 This is a structural diagram of a model evaluation device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] As described in the background technology, comparing the product images on the sales pages of this platform and the rival merchant platform is an important way to control the quality of the products sold, and the comparison work is performed by the image comparison model. Therefore, the performance of the image comparison model is extremely important, and the performance of the image comparison model needs to be evaluated frequently. At present, the evaluation of the performance of the image comparison model is performed manually. Specifically, the staff samples the results of the image comparison tasks that have been executed by the image comparison model, and then manually checks whether the comparison results of the image comparison model are accurate one by one to evaluate the performance of the image comparison model. However, the number of samples drawn by manual sampling is limited and it is inevitable that there will be some preference, resulting in sampling bias, which leads to deviations in the evaluation of the performance of the image comparison model. In addition, the workload of manually checking whether the comparison results of the image comparison model are accurate one by one is extremely large, and the work efficiency is low.

[0045] To solve the above problems, the present invention creatively proposes a model evaluation method, which automatically extracts execution result samples of the image comparison model through data tasks, sets accuracy and image recall rate indicators, and automatically and comprehensively judges the performance of the image comparison model, effectively avoiding the problem of small sampling number, reducing the workload of manual judgment and calculation, and the evaluation results are stable and reliable.

[0046] The model evaluation method, device and computer equipment provided by the present invention will be described in detail below in conjunction with specific embodiments.

[0047] Embodiment 1: This embodiment provides a model evaluation method for evaluating the performance of an image comparison model; Figure 1 As shown, the method includes:

[0048] S110, obtaining a sample set, the sample set including output data of the image comparison model; the output data of the image comparison model at least includes first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data.

[0049] Specifically, the picture comparison model compares two pictures at a time - the first picture and the second picture, the first picture and the second picture are pictures of goods sold on this platform and the rival platform, respectively. After comparing the first picture and the second picture, the picture comparison model outputs the first picture data, the second picture data and the matching value of the first picture data and the second picture data. The matching value of the first picture data and the second picture data is the overlap degree of the first picture and the second picture. Exemplarily, if the result of the picture comparison model comparing the first picture and the second picture is that the two pictures are exactly the same, then the matching value of the output first picture data and the second picture data is 100%; if the result of the picture comparison model comparing the first picture and the second picture is that the two pictures are completely different, then the matching value of the output first picture data and the second picture data is 0%. Of course, the matching value of the output first picture data and the second picture data can also be any value between 0%-100%, which is not elaborated in this embodiment.

[0050] In a preferred embodiment, obtaining the sample set includes:

[0051] S111, obtaining output data of the image comparison model within a preset time period.

[0052] Specifically, the image comparison model periodically completes the comparison task, and the sampling of the output results of the image comparison model is also periodic. To avoid duplication of work, only the output data of the newly added image comparison model is sampled. Exemplarily, the newly added output results of the image comparison model are obtained every 24 hours.

[0053] S112. Sampling the output data of the image comparison model within a preset time period to obtain a sample set.

[0054] More preferably, S112 includes:

[0055] S112-1. Sample the output data of the image comparison model within a preset time period to obtain sample data.

[0056] Specifically, the output data of the image comparison model within a preset time period is sampled by commodity category to obtain sampling data of each commodity category.

[0057] Exemplarily, the output data of the image comparison model within a preset time period is sampled according to the top category, shoe category, and skirt category respectively.

[0058] Of course, the sample set can also be obtained by sampling the newly added output data after the image comparison model iteration.

[0059] S112-2. Organize the sampled data into an EXCEL file. The sample set includes the EXCEL file.

[0060] Specifically, through the data task executed in advance, the online sampling data file, that is, the sampling result of the output data of the image comparison model within a preset time period, is imported through EXCEL and converted into test data to obtain a sample set.

[0061] In a preferred embodiment, before S120, the method further includes:

[0062] S113, filtering the output data of the image comparison model whose matching value between the first image data and the second image data in the sample set meets the first preset condition to obtain a first sub-sample set.

[0063] Specifically, in this embodiment, the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set is greater than or equal to 90% is screened to obtain the first sub-sample set.

[0064] S114, filtering the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the second preset condition to obtain a second sub-sample set.

[0065] Specifically, the output data of the image comparison model in which the matching value between the first image data and the second image data in the sample set is 100% is screened to obtain the second sub-sample set.

[0066] S120 , calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set.

[0067] In a preferred embodiment, this step includes:

[0068] S121. Query a database to obtain a barcode corresponding to each piece of first image data and second image data in a first sub-sample set.

[0069] S122. Obtaining a comparison accuracy rate of the image comparison model based on the barcodes corresponding to the first image data and the second image data.

[0070] Preferably, S122 includes:

[0071] S122-1. Perform domain judgment based on the barcodes corresponding to the first image data and the second image data.

[0072] If the barcodes corresponding to the first image data and the second image data are the same, then the image comparison model matches correctly;

[0073] S122-2. Count the number of correct matches of the image comparison model.

[0074] More preferably, the picture comparison model output data other than the picture comparison model matches determined to be correct by S122-1 in the first sub-sample set is manually verified, and the number of picture comparison model matches determined to be correct by S122-1 and the number of picture comparison model matches that are manually verified to be correct are counted.

[0075] More preferably, the judgment result data that S122-1 judges that the image comparison model matches correctly and the verification data that the manual verification result shows that the image comparison model matches correctly are saved in the database.

[0076] S122-3. Calculate the ratio of the number of correct matches of the image comparison model to the output data volume of the image comparison model in the first sub-sample set to obtain the comparison accuracy of the image comparison model.

[0077] Preferably, the overall accuracy is calculated and displayed by batch / date and by category on the sampling result interface.

[0078] S123. After a second preset time period, the second sub-sample set is sent to the image comparison model to obtain the number of recalled images, where the recalled images are the number of images whose matching values ​​with the first image data reach a preset threshold, obtained by the image comparison model from the current dynamic product image pool based on the first image data in the second sub-sample set.

[0079] Specifically, the output data of the image comparison model in the second sub-sample set is the output data that has been verified and confirmed. Exemplarily, the second sub-sample set includes the output data of the image comparison model whose matching value between the verified first image data and the second image data is 100%. The first image data in the second sub-sample set is pushed to the image comparison model through a kafka message. The image comparison model retrieves images with a matching value of 100% from the dynamic product image pool based on the first image. The number of images with a matching value of 100% retrieved from the dynamic product image pool based on the first image by the image comparison model is counted, i.e., the number of recalled images.

[0080] S124. Obtaining the image recall rate of the image comparison model based on the number of images whose matching values ​​with the first image data reach a preset threshold.

[0081] Preferably, S124 includes:

[0082] The image recall rate of the image comparison model is obtained by calculating the ratio of the number of images whose matching value with the first image data is greater than or equal to a preset threshold and the number of second image data in the second subsample set.

[0083] Specifically, the second sub-sample set includes the output data of the image comparison model whose matching value of the verified first image data and the second image data is 100%, that is, the second sub-sample set includes N pairs of completely identical first images and second images. Since the matching value of the first image and the second image is 100%, it can be determined that the dynamic product image pool, that is, the product images sold on this platform, must contain the same images. The image comparison model matches the images with a matching value of 100% in the dynamic product image pool data according to each first image data in the second sub-sample set, that is, the image comparison model retrieves the same image from the dynamic product image pool according to each first image. If the retrieval is successful, it is counted, and if the retrieval fails, it is not counted. Exemplarily, the second sub-sample set includes 100 verified output data of the image comparison model whose matching value with the second image data is 100%. The image comparison model matches the image data with a matching value of 100% in the dynamic product image pool data based on the 100 first image data, and the number of matches is 98. The number of images whose matching value with the first image data reaches a preset threshold is 98 divided by the total amount of first image data in the second sub-sample set, 100, and the image recall rate of the image comparison model is 98%.

[0084] S130, evaluating the performance of the image comparison model based on the comparison accuracy, image recall rate and preset evaluation rules.

[0085] Specifically, the performance of the image comparison model is evaluated by comprehensively comparing the comparison accuracy and the image recall rate. For example, the preset evaluation rule is: the performance of the image comparison model is qualified when the comparison accuracy is not less than 99% and the image recall rate is not less than 99%. After calculation, the comparison accuracy of the current image comparison model is 99% and the image recall rate is 98. The performance of the current image comparison model is unqualified and needs to be readjusted.

[0086] The model evaluation method provided in this embodiment is used to evaluate the performance of an image comparison model; the method includes: obtaining a sample set, the sample set including output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data; calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set; evaluating the performance of the image comparison model based on the comparison accuracy and the image recall rate; sampling and detecting the comparison results of the image comparison model, and evaluating the comparison accuracy and image recall rate indicators of the image comparison model according to the sampled sample set, thereby quickly, stably and accurately evaluating the performance of the image comparison model and reducing the workload of manual judgment and calculation.

[0087] Embodiment 2: This embodiment provides a model evaluation device for evaluating the performance of an image comparison model, the device comprising:

[0088] An acquisition module 21 is used to acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data;

[0089] A calculation module 22, used for calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set;

[0090] The evaluation module 23 is used to evaluate the performance of the image comparison model based on the comparison accuracy, the image recall rate and preset evaluation rules.

[0091] In a preferred embodiment, the acquisition module 21 includes:

[0092] A first acquisition unit 211 is used to acquire output data of the image comparison model within a preset time period;

[0093] The sampling unit 212 is used to sample the output data of the image comparison model within the preset time period to obtain a sample set.

[0094] More preferably, the sampling unit 212 includes:

[0095] The sampling subunit 2121 is used to sample the output data of the image comparison model within the preset time period to obtain sampled data;

[0096] The sorting subunit 2122 is used to sort the sampled data into an EXCEL file, and the sample set includes the EXCEL file.

[0097] More preferably, the sampling unit 212 is specifically configured to sample the output data of the image comparison model within the preset time period by commodity category to obtain sampling data of each commodity category.

[0098] Preferably, the device further comprises:

[0099] A first screening module 24, configured to screen the output data of the image comparison model whose matching value between the first image data and the second image data in the sample set meets a first preset condition to obtain a first sub-sample set;

[0100] The second screening module 25 is used to screen the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the second preset condition to obtain a second sub-sample set.

[0101] More preferably, the calculation module 22 includes:

[0102] A query unit 221 is used to query a database to obtain a barcode corresponding to each piece of first image data and second image data in the first sub-sample set;

[0103] A second acquisition unit 222, configured to obtain a comparison accuracy of the image comparison model based on the barcodes corresponding to the first image data and the second image data;

[0104] A third acquisition unit 223 is used to send the second sub-sample set to the image comparison model to obtain the number of recalled images after a second preset time period, where the recalled images are the number of images whose matching values ​​with the first image data reach a preset threshold, obtained by the image comparison model from the current dynamic product image pool according to the first image data in the second sub-sample set;

[0105] The fourth acquisition unit 224 is used to obtain the image recall rate of the image comparison model based on the number of images whose matching values ​​with the first image data reach a preset threshold.

[0106] More preferably, the second acquiring unit 222 includes:

[0107] The judgment subunit 2221 is used to perform domain judgment based on the barcodes corresponding to the first image data and the second image data; if the barcodes corresponding to the first image data and the second image data are the same, the image comparison model matches correctly;

[0108] A statistics subunit 2222 is used to count the number of correct matches of the image comparison model;

[0109] The first calculation subunit 2223 is used to calculate the ratio of the number of correct matches of the picture comparison model to the output data volume of the picture comparison model in the first sub-sample set to obtain the comparison accuracy of the picture comparison model.

[0110] More preferably, the fourth acquiring unit 224 includes:

[0111] The second calculation subunit 2241 is used to calculate the ratio of the number of pictures whose matching value with the first picture data is greater than or equal to a preset threshold and the number of the second picture data in the second sub-sample set to obtain the picture recall rate of the picture comparison model.

[0112] The model evaluation device provided in this embodiment is used to execute the model evaluation method provided in the first embodiment. Its beneficial effects are the same as those of the executed model evaluation method, and will not be elaborated here.

[0113] It should be noted that: the model evaluation device provided in the above embodiment only uses the division of the above functional modules as an example when triggering the model evaluation service. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the model evaluation device provided in the above embodiment and the embodiment of the model evaluation method provided in Example 1 belong to the same concept, that is, the device is based on the method, and its specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0114] Embodiment 3: The computer device provided in this embodiment includes a processor, a memory and a network interface connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an optimization method for an execution plan is implemented.

[0115] Those skilled in the art will appreciate that the structure in this embodiment is merely a partial structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0116] As a preferred implementation, in an embodiment of the present invention, a computer device includes a memory and a processor, the memory stores a computer program that can be run on the processor, and the processor implements the following steps when executing the computer program:

[0117] Acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data;

[0118] Calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set;

[0119] The image comparison model performance is evaluated based on the comparison accuracy, the image recall rate and preset evaluation rules.

[0120] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0121] Of course, the above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any modifications made according to the spirit of the main technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A model evaluation method, characterized in that: Used to evaluate the performance of an image comparison model; the method comprises: Acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data; Calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set; Evaluating the performance of the image comparison model based on the comparison accuracy, the image recall rate and preset evaluation rules; Before calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set, the method further includes: Filter the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the first preset condition to obtain a first sub-sample set; Filter the output data of the picture comparison model whose matching value between the first picture data and the second picture data in the sample set meets the second preset condition to obtain a second sub-sample set; The calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set includes: Querying a database to obtain a barcode corresponding to each piece of first image data and second image data in the first sub-sample set; Obtaining a comparison accuracy rate of the image comparison model based on the barcodes corresponding to the first image data and the second image data; After a second preset time period, the second sub-sample set is sent to the image comparison model to obtain the number of recalled images, where the recalled images are the number of images whose matching values ​​with the first image data reach a preset threshold, obtained by the image comparison model from the current dynamic product image pool according to the first image data in the second sub-sample set; The image recall rate of the image comparison model is obtained based on the number of images whose matching values ​​with the first image data reach a preset threshold.

2. The model evaluation method according to claim 1, characterized in that: The obtaining of the sample set comprises: Obtaining output data of the image comparison model within a preset time period; The output data of the image comparison model within the preset time period is sampled to obtain a sample set.

3. The model evaluation method according to claim 2, characterized in that: The sampling of the output data of the image comparison model within the preset time period to obtain a sample set includes: Sampling the output data of the image comparison model within the preset time period to obtain sample data; The sampled data is organized into an EXCEL file, and the sample set includes the EXCEL file.

4. The model evaluation method according to claim 2 or 3, characterized in that: The sampling of the output data of the image comparison model within the preset time period to obtain the sampled data includes: The output data of the image comparison model within the preset time period is sampled by commodity category to obtain sampling data of each commodity category.

5. The model evaluation method according to claim 1, characterized in that: The obtaining of the comparison accuracy of the image comparison model based on the barcodes corresponding to the first image data and the second image data includes: Performing domain determination based on the barcodes corresponding to the first image data and the second image data; If the barcodes corresponding to the first image data and the second image data are the same, then the image comparison model matches correctly; Counting the number of correct matches of the image comparison model; The comparison accuracy of the picture comparison model is obtained by calculating the ratio of the number of correct matches of the picture comparison model to the output data volume of the picture comparison model in the first sub-sample set.

6. The model evaluation method according to claim 1, characterized in that: The obtaining of the image recall rate of the image comparison model based on the number of images whose matching values ​​with the first image data reach a preset threshold comprises: The image recall rate of the image comparison model is obtained by calculating the ratio of the number of images whose matching values ​​with the first image data are greater than or equal to a preset threshold and the number of the second image data in the second sub-sample set.

7. A model evaluation device for implementing the model evaluation method according to claim 1, characterized in that: For evaluating the performance of an image comparison model, the device comprises: an acquisition module, configured to acquire a sample set, wherein the sample set includes output data of the image comparison model; the output data of the image comparison model includes at least first image data, second image data associated with the first image data, and a matching value between the first image data and the second image data; A calculation module, used for calculating the comparison accuracy of the image comparison model and the image recall rate of the image comparison model based on the sample set; An evaluation module is used to evaluate the performance of the image comparison model based on the comparison accuracy, the image recall rate and preset evaluation rules.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images