Video detection algorithm result analysis method and system and storage medium

Abstract

The invention discloses a video detection algorithm result analysis method and system and a storage medium. The method comprises the steps of obtaining a detection result video and a detection result image; automatically classifying the detection result image according to a preset confidence threshold to obtain a first image of which the confidence value is greater than or equal to the preset confidence threshold and a second image of which the confidence value is smaller than the preset confidence threshold, and adding a first label to the first image; obtaining a first timestamp of each detection result image or the second image, a frame number of a corresponding video frame in the detection result video and a second timestamp of the corresponding video frame, and establishing an association relationship between the video frame and the detection result image or establishing an association relationship between the video frame and the second image; displaying the second image in an image verification area, and displaying a detection result video in a video playing area; verifying the second image based on the association relationship and the target ID, and adding a second label to the second image; and automatically generating an analysis report based on the first label and the second label.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video detection algorithm analysis, in particular to a video detection algorithm result analysis method and system and a storage medium. BACKGROUND

[0002] Video target detection not only needs to consider the position of the target in space, but also needs to process the change in the time dimension to identify the motion trajectory of the target in different frames, so the test of the video target detection algorithm is more complex.

[0003] At present, the analysis of the video detection result mainly depends on the verification personnel to view the detection result images one by one to judge whether the target detected by the algorithm has any problem, and then the verification personnel analyzes and classifies the detection result images with problems and the reasons for the problems. The whole analysis process needs the verification personnel to switch between the image viewing software and the video playing software, viewing the video while searching for the pictures at the same time, which has the problems of complicated operation and low analysis efficiency. SUMMARY

[0004] To solve the above technical problems, the present application provides a video detection algorithm result analysis method, which comprises the following steps: obtaining a detection result video and a detection result image, the detection result image comprising at least an image name, a first timestamp, a target ID and a confidence value; automatically classifying the detection result image according to a pre-set confidence threshold to obtain a first image with a confidence value greater than or equal to the pre-set confidence threshold and a second image with a confidence value less than the pre-set confidence threshold, and adding a first label to the first image; obtaining the first timestamp of each detection result image or the second image, the frame number of the corresponding video frame in the detection result video and the second timestamp of the corresponding video frame, establishing the association between the video frame and the detection result image or the association between the video frame and the second image; displaying the second image in an image verification area and displaying the detection result video in a video playing area; verifying the second image based on the association and the target ID, and adding a second label to the second image; generating an analysis report automatically based on the first label and the second label.

[0005] Optionally, obtaining the first timestamp of each detection result image or the second image, the frame number of the corresponding video frame in the detection result video and the second timestamp of the corresponding video frame, establishing the association between the video frame and the detection result image or the association between the video frame and the second image, at least comprises the following steps: acquiring a first timestamp of each detection result image or second image, a frame number of a corresponding video frame in the detection result video, and a second timestamp of the corresponding video frame; storing the second timestamp and the frame number in the form of a hash table; matching the first timestamp and the second timestamp to obtain the frame number of the corresponding video frame of the detection result image or the second image; acquiring an image name of each detection result image or second image, and establishing an association between the video frame and the detection result image or between the video frame and the second image based on the image name and the frame number.

[0006] Optionally, the second image is verified based on the association and the target ID, and a second label is added to the second image, including at least the following steps: acquiring a second image selection instruction; based on the association, acquiring the frame number of the corresponding video frame of the selected second image, and positioning the current video frame of the video playing area to the corresponding video frame; adding a second label to the second image based on the target ID and a preset label type; the preset label type at least includes correct capture, incorrect capture, repeated capture, and non-optimal capture, and the second label is one of the preset label types.

[0007] Optionally, after adding the second label to the second image, the following steps are further included: determining whether the second label includes one of incorrect capture, repeated capture, and non-optimal capture, and if so, the following steps are executed: acquiring a video screenshot according to a video screenshot acquisition instruction; acquiring a corresponding video clip according to a video clip acquisition instruction; acquiring the error cause of the second image; adding the second label to the video screenshot and the video clip, and saving the second label in association with the error cause.

[0008] Optionally, the analysis report at least includes a video detection result correctness rate, an error rate, an error type, and a cause distribution chart.

[0009] Optionally, based on the first label and the second label, an analysis report is automatically generated, including at least the following steps: respectively counting the number of the first label and the second label, and calculating the video detection result correctness rate and the video detection result error rate; acquiring the second label including one of incorrect capture, repeated capture, and non-optimal capture, and the corresponding error cause, and generating a video detection result error type and cause distribution chart.

[0010] Optionally, during the verification process, real-time record the verification progress data, the verification progress data at least includes the frame number of the current video frame of the video playing area and the second image currently displayed in the image verification area.

[0011] Optionally, during the verification process, also real-time detect whether there is a pause analysis instruction; If there is a pause analysis instruction, save the current verification progress data, and real-time detect whether there is a restart analysis instruction; If there is a restart analysis instruction, read and automatically position the detection result video to the corresponding video frame according to the saved verification progress data, and display the corresponding second image according to the association relationship.

[0012] Corresponding to the video detection algorithm result analysis method, the application provides a video detection algorithm result analysis system, which comprises: A data acquisition module is configured to acquire a detection result video and a detection result image, wherein the detection result image at least includes an image name, a first timestamp, a target ID and a confidence value; A preprocessing module is configured to automatically classify the detection result image according to a pre-set confidence threshold, to obtain a first image with a confidence value greater than or equal to the pre-set confidence threshold and a second image with a confidence value less than the pre-set confidence threshold, and to add a first label to the first image; An association relationship establishing module is configured to acquire the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video and the second timestamp of the corresponding video frame, to establish the association relationship between the video frame and the detection result image or the association relationship between the video frame and the second image; A display module is configured to display the second image in an image verification area and the detection result video in a video playing area; A verification module is configured to verify the second image based on the association relationship and the target ID, and to add a second label to the second image; A report generation module is configured to automatically generate an analysis report based on the first label and the second label.

[0013] In addition, to achieve the above-mentioned purposes, the application further provides a computer readable storage medium, wherein the computer readable storage medium stores a video detection algorithm result analysis program, and the video detection algorithm result analysis program is executed by a processor to realize the steps of the video detection algorithm result analysis method as described above.

[0014] The application realizes efficient analysis of video detection results, effectively improves analysis efficiency and accuracy.

[0015] The application can quickly and accurately obtain the frame number of the video frame corresponding to the detection result image by storing the second timestamp and frame number in a hash table and establishing an association relationship by timestamp matching, provides a reliable basis for subsequent image verification and analysis report generation, ensures the smoothness and timeliness of the entire analysis process, and further improves the quality and speed of video detection algorithm result analysis.

[0016] The application realizes that clicking the second image can jump to the corresponding progress position in the video, without manually dragging the video playback progress bar to find the position of the second image in the video, greatly improving the operation efficiency of the verifier. At the same time, the second label is added according to the target ID and the preset label type, so that the verifier can intuitively and conveniently verify the second image and mark detailed information; in addition, the preset multiple label types refine the classification of the second image, which helps to more accurately understand various problems in the detection result, provides rich information support for subsequent targeted improvement of the video detection algorithm, and enhances the depth and professionalism of the analysis.

[0017] After adding the second label, the application judges whether it contains a specific problem type, and if so, further obtains a video screenshot, a video segment and an error reason, and saves them in association. This not only enriches the analysis data, but also more comprehensively records and presents the problems in video detection, facilitates subsequent in-depth analysis of the causes and effects of the problems, provides a more sufficient basis for optimizing the video detection algorithm, and also facilitates tracking and backtracking of the problems, further perfecting the process and content of the video detection algorithm result analysis.

[0018] The application automatically generates an analysis report, so that the verifier can quickly understand the overall quality and problem distribution of the video detection, which is conducive to timely discovering and positioning problems, thereby more effectively guiding the improvement and optimization of the video detection algorithm, and improving the readability and practicality of the analysis result.

[0019] The application calculates the accuracy and error rate by counting the number of labels, generates a distribution chart by obtaining a specific second label and an error reason, further refines the generation process of the analysis report, and can more accurately reflect various indicators and problems of the video detection result.

[0020] The application can ensure the continuity and traceability of the analysis process by recording the verification progress data in real time during the verification process, including the video frame number and the displayed second image. Even if there is an unexpected interruption during the analysis process, the current verification progress can be accurately known, avoiding the risk of repeated work and data loss, improving the reliability and stability of the entire analysis process, and providing a strong guarantee for efficient completion of video detection algorithm result analysis.

[0021] The application greatly enhances the flexible control ability of the system on the analysis process by detecting the pause analysis instruction in real time and saving the progress data when pausing, and positioning the video and the displayed image according to the progress data when resuming. This allows the verifier to pause and resume the analysis work at any time according to actual needs, improving the convenience and flexibility of the analysis operation, while ensuring the continuity and integrity of the analysis process, further improving the user experience and work efficiency of the video detection algorithm result analysis system. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings, which are included to provide a further understanding of the application, form a part of the application and help to explain the application illustrate the specific embodiments of the application and its description, and do not constitute an improper limitation on the application. In the drawings: Figure 1 Flowchart of an embodiment of the video detection algorithm result analysis method of the application; Figure 2 Frame diagram of an embodiment of the video detection algorithm result analysis system of the application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme of the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, not all the embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and do not limit the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0024] As shown in Figure 1 A video detection algorithm result analysis method of the application includes the following steps: Obtain the detection result video and the detection result image, and the detection result image includes at least image name, first timestamp, target ID and confidence value; According to the preset confidence threshold, the detection result image is automatically classified to obtain a first image with a confidence value greater than or equal to the preset confidence threshold and a second image with a confidence value less than the preset confidence threshold, and a first label is added to the first image; A first timestamp of each detection result image or the second image, a frame number of a corresponding video frame in the detection result video, and a second timestamp of the corresponding video frame are obtained, and an association relationship between the video frame and the detection result image or an association relationship between the video frame and the second image is established; The second image is displayed in an image verification area, and the detection result video is displayed in a video playing area; The second image is verified based on the association relationship and the target ID, and a second label is added to the second image; Based on the first label and the second label, an analysis report is automatically generated.

[0025] In the embodiment, each video frame in the detection result video includes a target detection tracking number, a second timestamp, and a target ID, and the detection result video is used to check the detection result image, to verify whether the target ID detected by the detection result image is correct.

[0026] Preferably, the detection result image can be obtained by the output of any existing video detection model, and the analysis of the video detection result is not limited to the source of the detection result image.

[0027] It should be noted that the first label is a "default correct" label, so the first image does not need to be verified again, and the automatic confirmation of the first image can be realized by the preset confidence threshold.

[0028] It should be noted that the image verification area and the video playing area are located in different areas of the same interface, and the verification work of the second image can be completed in one interface, thereby greatly improving the verification efficiency.

[0029] The present application obtains the detection result video and the detection result image, and automatically classifies according to the preset confidence threshold, directly adds a first label to the first image with high confidence, establishes an association relationship between the video frame and the detection result image, places the second image in the image verification area for verification and adds a second label, and finally generates an analysis report based on the two types of labels; the analysis of the video detection result is realized, and the analysis efficiency and accuracy are effectively improved.

[0030] In the embodiment, a first timestamp of each detection result image or the second image, a frame number of a corresponding video frame in the detection result video, and a second timestamp of the corresponding video frame are obtained, and an association relationship between the video frame and the detection result image or an association relationship between the video frame and the second image is established, at least including the following steps: acquire a first timestamp of each detection result image or second image, a frame number of a corresponding video frame in the detection result video, and a second timestamp of the corresponding video frame; the corresponding video frame is a video frame corresponding to each detection result image or a video frame corresponding to each second image; store the second timestamp and the frame number in the form of a hash table; match the first timestamp and the second timestamp to acquire the frame number of the video frame corresponding to the detection result image or the second image; acquire an image name of each detection result image or second image, and establish an association between the video frame and the detection result image or between the video frame and the second image based on the image name and the frame number.

[0031] It should be noted that the detection result image includes the second image, but since the present application only needs to verify the second image below the confidence threshold, the association can be established only for the second image, and the association between the video frame and the second image is established. Alternatively, for the convenience of operation, the association between the video frame and the detection result image can also be established for the detection result image.

[0032] Preferably, the first timestamp of the detection result image or the second image is acquired by an OCR text recognition technology, and the second timestamp of the detection result video is acquired. The detection result image and the video frame time sequence are associated or the second image and the video frame time sequence are associated by comparing whether the time cut is the same, and the association is saved in the form of a hash table {image name: video frame number} in the disk.

[0033] The present application can quickly and accurately acquire the frame number of the video frame corresponding to the detection result image by storing the second timestamp and the frame number in the form of a hash table and matching the timestamps to establish the association, which provides a reliable basis for subsequent image verification and analysis report generation, ensures the smoothness and timeliness of the entire analysis process, and further improves the quality and speed of the video detection algorithm result analysis.

[0034] In the present embodiment, the second image is verified based on the association and the target ID, and a second label is added to the second image, including at least the following steps: acquire a second image selection instruction; based on the association, acquire the frame number of the video frame corresponding to the selected second image, and position the current video frame of the video playing area to the corresponding video frame; based on the target ID and a preset label type, add a second label to the second image; the preset label type includes at least correct capture, incorrect capture, repeated capture, and non-optimal capture, and the second label is one of the preset label types.

[0035] In the present embodiment, the specific meanings of each preset label type are as follows: True Positive (TP): The algorithm successfully detects the target, and the angle, position, clarity, etc. of the target image is the best among the multiple frames containing the target.

[0036] False Positive (FP): The algorithm incorrectly detects the target, and the target image is not actually the expected target.

[0037] Duplicate Detection: The algorithm detects multiple targets for the same target, and the multiple target images are actually the same target.

[0038] Suboptimal Detection: The algorithm detects the target, but the clarity or position of the target image is not ideal.

[0039] It can be understood that since there can be multiple target IDs in the same video frame, the target ID of the second image needs to be used to determine which ID in the detection result video each target ID corresponds to during verification.

[0040] Specifically, the verifier can generate a second image selection instruction by clicking a second image in the image verification area.

[0041] The present application uses the association relationship to locate the video playback area to the corresponding video frame by obtaining the second image selection instruction, enabling the second image to jump to the corresponding progress position in the video by clicking, without manually dragging the video playback progress bar to find the position of the second image in the video, greatly improving the operation efficiency of the verifier. At the same time, the second label is added according to the target ID and the preset label type, enabling the verifier to intuitively and conveniently verify and label detailed information for the second image; in addition, the preset multiple label types refine the classification of the second image, which helps to more accurately understand various problems in the detection result, provides rich information support for subsequent targeted improvement of the video detection algorithm, and enhances the depth and professionalism of the analysis.

[0042] In this embodiment, after adding the second label to the second image, the following steps are further included: determining whether the second label includes one of false detection, duplicate detection, and suboptimal detection, and if so, performing the following steps: obtaining a video screenshot according to the video screenshot obtaining instruction; obtaining a corresponding video segment according to the video clip instruction; obtaining the error reason of the second image; adding the second label to the video screenshot and the video segment, and saving the second label in association with the error reason.

[0043] In the embodiment, the video screenshot acquisition instruction can be triggered by user operation or automatically triggered.

[0044] Preferably, the verifier can actively input the error reason so that the system obtains the error reason of the second image.

[0045] After adding the second label, the application determines whether it contains a specific problem type, and if so, further obtains a video screenshot, a video segment, and an error reason, and performs associated saving. This not only enriches the analysis data, but also more comprehensively records and presents the problems in video detection, facilitates subsequent in-depth analysis of the causes and effects of the problems, provides a more sufficient basis for optimizing the video detection algorithm, and also facilitates tracking and backtracking of the problems, further perfecting the process and content of video detection algorithm result analysis.

[0046] In the embodiment, the analysis report at least includes a video detection result accuracy rate, an error rate, an error type, and a reason distribution chart.

[0047] The application automatically generates an analysis report, so that the verifier can quickly understand the overall quality and problem distribution of the video detection, which is beneficial to timely discovering and positioning the problems, thereby more effectively guiding the improvement and optimization of the video detection algorithm, and improving the readability and practicality of the analysis result.

[0048] In the embodiment, based on the first label and the second label, an analysis report is automatically generated, including at least the following steps: The number of the first label and the second label is respectively counted, and the video detection result accuracy rate and the video detection result error rate are calculated; The second label including one of the error snapshot, the repeated snapshot, and the non-optimal snapshot, and the corresponding error reason are obtained, and a video detection result error type and reason distribution chart is generated.

[0049] The application calculates the accuracy rate and the error rate by counting the number of labels, obtains specific second labels and error reasons to generate a distribution chart, further refines the generation process of the analysis report, and can more accurately reflect various indicators and problems of the video detection result.

[0050] In the embodiment, during the verification process, verification progress data is recorded in real time, and the verification progress data at least includes a frame number of a current video frame of a video playing area and a second image currently displayed in an image verification area.

[0051] The application can ensure the continuity and traceability of the analysis process by recording the verification progress data, including the video frame number and the displayed second image, in real time during the verification process. Even if there is an unexpected interruption during the analysis process, the current verification progress can be accurately known, avoiding the risk of repeated work and data loss, improving the reliability and stability of the entire analysis process, and providing a strong guarantee for efficient completion of video detection algorithm result analysis.

[0052] In this embodiment, during the verification process, it is also detected in real time whether there is a pause analysis instruction; If there is a pause analysis instruction, the current verification progress data is saved, and it is detected in real time whether there is a restart analysis instruction; If there is a restart analysis instruction, the detection result video is automatically positioned to the corresponding video frame according to the saved verification progress data, and the corresponding second image is displayed according to the association relationship.

[0053] The application greatly enhances the flexible control ability of the system on the analysis process by detecting the pause analysis instruction in real time and saving the progress data when pausing, and positioning the video and displaying the image according to the progress data when restarting. This allows the verifier to pause and resume the analysis work at any time according to the actual needs, improving the convenience and flexibility of the analysis operation, while also ensuring the continuity and integrity of the analysis process, further improving the user experience and work efficiency of the video detection algorithm result analysis system.

[0054] As shown in Figure 2 The application also provides a video detection algorithm result analysis system, which includes: A data acquisition module 10 is configured to acquire a detection result video and a detection result image, and the detection result image includes at least an image name, a first timestamp, a target ID, and a confidence value. A preprocessing module 20 is configured to automatically classify the detection result image according to a preset confidence threshold, obtain a first image with a confidence value greater than or equal to the preset confidence threshold and a second image with a confidence value less than the preset confidence threshold, and add a first label to the first image. An association relationship establishment module 30 is configured to acquire the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video, and the second timestamp of the corresponding video frame, and establish the association relationship between the video frame and the detection result image or the association relationship between the video frame and the second image. A display module 40 is configured to display the second image in an image verification area and display the detection result video in a video playback area. A verification module 50 is configured to verify the second image based on the association relationship and the target ID, and add a second label to the second image. The report generation module 60 is configured to automatically generate an analysis report based on the first label and the second label.

[0055] Preferably, the verification module 50 is further configured to, after adding the second label to the second image, determine whether the second label comprises one of an error snapshot, a repeated snapshot and a non-optimal snapshot, and if so, perform the following steps: According to the video screenshot acquisition instruction, acquire a video screenshot; According to the video clip acquisition instruction, acquire a corresponding video clip; Acquire an error cause of the second image; Add the second label to the video screenshot and the video clip, and save the second label in association with the error cause.

[0056] During the verification process, the verification module 50 is further configured to record verification progress data in real time, the verification progress data at least comprising a frame number of a current video frame of the video playing area and a second image currently displayed in the image verification area; and detect in real time whether a pause analysis instruction exists; if the pause analysis instruction exists, save the current verification progress data, and detect in real time whether a restart analysis instruction exists; if the restart analysis instruction exists, read and automatically position the detection result video to a corresponding video frame according to the saved verification progress data, and display a corresponding second image according to the association relationship.

[0057] The embodiment of the present application further provides a computer readable storage medium, which can be a computer readable storage medium contained in the memory in the above-mentioned embodiment, or can exist independently and not be assembled into the device. The computer readable storage medium stores at least one instruction, which is loaded and executed by the processor to realize the video detection algorithm result analysis method shown in the embodiment of the present application. Figure 1 The computer readable storage medium can be a read-only memory, a disk or an optical disk.

[0058] It should be noted that each embodiment in the present specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to. For the system embodiment and the storage medium embodiment, since they are basically similar to the method embodiment, they are described more simply, and the relevant parts can be referred to the part of the method embodiment.

[0059] Also, as used herein, the terms "comprise", "comprising", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0060] The foregoing description shows and describes preferred embodiments of the application, but it is to be understood that the application is not limited to the particular examples, arrangements, or methodologies described, and that various other embodiments, modifications, and changes can be made thereto without departing from the spirit and scope of the application as set forth in the following claims.

Claims

1. A method for analyzing the results of a video detection algorithm, characterized in that, Includes the following steps: Acquire detection result videos and images. Detection result images must include at least the image name, first timestamp, target ID, and confidence value. The detection result images are automatically classified according to a preset confidence threshold to obtain a first image with a confidence value greater than or equal to the preset confidence threshold and a second image with a confidence value less than the preset confidence threshold, and a first label is added to the first image; Obtain the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video, and the second timestamp of the corresponding video frame, and establish the association between the video frame and the detection result image or establish the association between the video frame and the second image. The second image is displayed in the image verification area, and the detection result video is displayed in the video playback area; The second image is verified based on the association relationship and the target ID, and a second tag is added to the second image; An analysis report is automatically generated based on the first and second tags.

2. The video detection algorithm result analysis method according to claim 1, characterized in that, Obtain the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video, and the second timestamp of the corresponding video frame; establish the association between the video frame and the detection result image or establish the association between the video frame and the second image, including at least the following steps: Obtain the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video, and the second timestamp of the corresponding video frame; The second timestamp and frame number are stored in the form of a hash table; Match the first timestamp with the second timestamp to obtain the frame number of the video frame corresponding to the detection result image or the second image; Obtain the image name of each detection result image or second image, and establish the association between the video frame and the detection result image or the association between the video frame and the second image based on the image name and frame number.

3. The video detection algorithm result analysis method according to claim 1, characterized in that, Verify the second image based on the association relationship and target ID, and add a second tag to the second image, including at least the following steps: Obtain the second image selection instruction; Based on the association relationship, obtain the frame number of the video frame corresponding to the selected second image, and locate the current video frame in the video playback area to the corresponding video frame. A second label is added to the second image based on the target ID and the preset label type. The preset label types include at least correct capture, incorrect capture, repeated capture, and non-optimal capture, and the second label is one of the preset label types.

4. The video detection algorithm result analysis method according to claim 3, characterized in that, After adding a second label to the second image, it also includes: Determine if the second label includes one of the following: erroneous capture, duplicate capture, or suboptimal capture. If so, proceed with the following steps: Obtain video screenshots based on the video screenshot acquisition command; Obtain the corresponding video segment according to the video extraction command; The reason for the error in obtaining the second image; Add a second tag to the video screenshot and video clip, and save it along with the error reason.

5. The video detection algorithm result analysis method according to claim 4, characterized in that, The analysis report should include at least a chart showing the accuracy rate, error rate, error types, and causes of the video detection results.

6. The video detection algorithm result analysis method according to claim 5, characterized in that, Based on the first and second tags, an analysis report is automatically generated, including at least the following steps: Count the number of the first label and the second label respectively, and calculate the accuracy and error rate of the video detection results. Obtain a second label, including one of the following: erroneous capture, duplicate capture, and non-optimal capture, along with the corresponding error reason, and generate a chart showing the distribution of error types and reasons in the video detection results.

7. The video detection algorithm result analysis method according to claim 1, characterized in that, During the verification process, the verification progress data is recorded in real time. The verification progress data includes at least the frame number of the current video frame in the video playback area and the second image currently displayed in the image verification area.

8. The video detection algorithm result analysis method according to claim 7, characterized in that, During the verification process, it also detects in real time whether there is a command to pause analysis; If a pause analysis command is issued, the current verification progress data is saved, and a restart analysis command is detected in real time. If a restart analysis command is issued, the system reads and automatically locates the detection result video to the corresponding video frame based on the saved verification progress data, and displays the corresponding second image according to the correlation.

9. A video detection algorithm result analysis system, characterized in that, include: The data acquisition module is used to acquire detection result videos and detection result images. The detection result images include at least the image name, first timestamp, target ID, and confidence value. The preprocessing module is used to automatically classify the detection result images according to a preset confidence threshold, to obtain a first image with a confidence value greater than or equal to the preset confidence threshold and a second image with a confidence value less than the preset confidence threshold, and to add a first label to the first image; The association establishment module is used to obtain the first timestamp of each detection result image or second image, the frame number of the corresponding video frame in the detection result video and the second timestamp of the corresponding video frame, and to establish the association between the video frame and the detection result image or the association between the video frame and the second image. The display module is used to display the second image in the image verification area and the detection result video in the video playback area; The verification module verifies the second image based on the association relationship and the target ID, and adds a second tag to the second image; The report generation module is used to automatically generate analysis reports based on the first and second tags.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a video detection algorithm result analysis program, which, when executed by a processor, implements the steps of the video detection algorithm result analysis method as described in any one of claims 1 to 8.

Images