A method, device, equipment and storage medium for updating target detection model

Abstract

The present invention discloses a method, device, equipment and storage medium for updating a target detection model. The target detection model is used to perform target detection, obtain the confidence of the current target detection, and judge the confidence of the current target detection with a preset threshold; when it is judged that the confidence of the current target detection is less than the preset threshold, the camera is controlled to collect and store the image data of the current frame, and the collected image data reaching a preset number are manually annotated; the manually annotated image data is fused with historical image data to generate fused image data; the fused image data is compared and studied with the historical image data to obtain the key-value pairs of the model parameters, and the target detection model is updated according to the key-value pairs. Compared with the prior art, the present invention improves the robustness of the model by fusing and comparing the collected image data with the historical image data, so that the classification of difficult samples is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robots, and in particular to a method, device, equipment and storage medium for updating a target detection model. Background Art

[0002] Object detection is to find all objects of interest in the image, including two subtasks: object location and object classification, and determine the category and location of the object. The process of object detection based on deep learning mainly consists of three steps: first, data collectors use professional cameras to shoot the target in real scenes to obtain raw data; second, the deep learning model deployed on the robot extracts features from the input data through trained logical reasoning functions; third, after obtaining the features proposed in the second step, the deep learning model continues to classify and judge the features extracted from the input data, and then identify the target in the input data image, and give the probability that the target belongs to a category.

[0003] Feature extraction based on deep learning can use the computing power of computers to extract features. Object detection based on deep learning is more convenient than traditional methods. However, the data encountered in practice is often not ideal data. In the field of object detection, long-tail data distribution or data label mislabeling are often encountered. One of the existing technologies is to improve the loss function and make the loss weight of the majority class higher. This can improve the accuracy of classification and improve the problem of class imbalance, but it cannot solve the problem of mislabeling or difficult sample classification. Another existing technology is to expand the data using data enhancement strategies. This can not only solve the problem of class imbalance, but also solve the problem of difficult sample classification and mislabeling. However, the existing data enhancement methods are either geometric transformation or random cropping. Although these methods can improve the accuracy of classification, they will lose some information such as background information, local object information, etc., resulting in fewer technologies available for difficult sample classification and more difficult implementation. Moreover, changing the loss function will distort the representation of the data. When the data does not have a class imbalance problem, the loss function needs to be changed, and repeated deployment of the model will cause the update process to be too complicated. Summary of the invention

[0004] The technical problem to be solved by the present invention is: to provide a method, device, equipment and storage medium for updating a target detection model, by fusing and comparing the collected image data with the historical image data, the robustness of the model is improved, and the classification of difficult samples is made more accurate.

[0005] In order to solve the above technical problems, the present invention provides a method for updating a target detection model, comprising:

[0006] Performing target detection through a target detection model, obtaining the confidence of the current target detection, and comparing the confidence of the current target detection with a preset threshold;

[0007] When it is determined that the confidence level of the current target detection is less than a preset threshold, the camera is controlled to collect and store image data of the current frame, and the collected image data reaching a preset amount are manually annotated;

[0008] Fusing the manually annotated image data with historical image data to generate fused image data;

[0009] The fused image data is compared and learned with the historical image data to obtain key-value pairs of model parameters, and the target detection model is updated according to the key-value pairs.

[0010] Furthermore, manually labeling the image data that has been collected and reaches a preset amount is specifically:

[0011] Acquiring the storage amount of the image data in real time, comparing the storage amount with a preset amount, and when the storage amount is greater than or equal to the preset amount, sending the image data reaching the preset amount to a server side so as to manually annotate the image data;

[0012] When the storage amount is less than the preset amount, wait for the next storage and update the current storage amount in real time.

[0013] Furthermore, the image data after manual annotation is fused with the historical image data to generate fused image data, specifically:

[0014] Selecting a key part of the manually annotated image data, and cropping the key part to obtain key part data;

[0015] The key part data is fused into the historical image data of the same type to generate fused image data.

[0016] Furthermore, the comparing and learning of the fused image data and the historical image data is specifically as follows:

[0017] When the target detection model is trained, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are of the same type of data is output.

[0018] Furthermore, the present invention provides an updating device for a target detection model, comprising: a confidence acquisition module, an image data acquisition module, an image fusion module and a comparison module;

[0019] The confidence acquisition module is used to perform target detection through the target detection model, obtain the confidence of the current target detection, and compare the confidence of the current target detection with a preset threshold;

[0020] The image data acquisition module is used to control the camera to acquire and store the image data of the current frame when it is determined that the confidence level of the current target detection is less than a preset threshold, and to manually annotate the acquired image data reaching a preset amount;

[0021] The image fusion module is used to fuse the manually annotated image data with historical image data to generate fused image data;

[0022] The comparison module is used to compare and learn the fused image data with the historical image data to obtain key-value pairs of model parameters, and update the target detection model according to the key-value pairs.

[0023] Furthermore, the image data acquisition module is used to manually annotate the image data that has been acquired and reaches a preset amount, specifically:

[0024] Acquiring the storage amount of the image data in real time, comparing the storage amount with a preset amount, and when the storage amount is greater than or equal to the preset amount, sending the image data reaching the preset amount to a server side so as to manually annotate the image data;

[0025] When the storage amount is less than the preset amount, wait for the next storage and update the current storage amount in real time.

[0026] Furthermore, the image fusion module is used to fuse the manually annotated image data with the historical image data to generate fused image data, specifically:

[0027] Selecting a key part of the manually annotated image data, and cropping the key part to obtain key part data;

[0028] The key part data is fused into the historical image data of the same type to generate fused image data.

[0029] Furthermore, the comparison module is used to compare and learn the fused image data with the historical image data, specifically:

[0030] When the target detection model is trained, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are of the same type of data is output.

[0031] Furthermore, the present invention provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a method for updating a target detection model as described in any one of the above when executing the computer program.

[0032] Furthermore, the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a target detection model updating method as described in any one of the above.

[0033] Compared with the prior art, the target detection model updating method, device, equipment and storage medium of the embodiment of the present invention have the following beneficial effects:

[0034] When executing the target detection instruction, the confidence of the current target detection is judged against the preset threshold through the target detection model; when it is judged that the confidence of the current target detection is less than the preset threshold, the camera is controlled to collect and store the image data of the current frame, so that the camera only collects image data with poor confidence. Since image data with poor confidence is often misclassified, the present invention also manually annotates the stored image data with poor confidence to correctly classify it, and fuses the manually annotated image data with historical image data, and then compares and learns the two to obtain the key-value pairs of model parameters, and updates the target detection model according to the key-value pairs. Compared with the prior art, the present invention enhances the accuracy of the model in classifying difficult samples and improves the robustness of the model by fusing and comparing the collected image data with the historical image data. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of an embodiment of a method for updating a target detection model provided by the present invention;

[0036] Figure 2 It is a structural schematic diagram of an embodiment of a device for updating a target detection model provided by the present invention;

[0037] Figure 3 It is a schematic diagram of image fusion of an embodiment of a method for updating a target detection model provided by the present invention;

[0038] Figure 4 It is a schematic diagram of image contrast learning of an embodiment of a method for updating a target detection model provided by the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Example 1

[0041] See also Figure 1 , Figure 1 FIG. 1 is a flow chart of an embodiment of a method for updating a target detection model provided by the present invention. Figure 1 As shown, the method includes steps 101 to 104, which are specifically as follows:

[0042] Step 101: Perform target detection through a target detection model, obtain the confidence of the current target detection, and compare the confidence of the current target detection with a preset threshold.

[0043] In this embodiment, the target detection model is set in the robot, and the target detection instruction is issued to the robot so that the communication module in the robot system receives the target detection instruction; wherein, the robot system includes a camera module, a data storage module and a central control module in addition to the communication module and the target detection model; specifically, the camera module is used to receive the instruction of the central control module to complete the setting of the height, inclination and other parameters of the camera device; obtain image data, and after the target detection model completes the target detection and recognition, the central control module issues an instruction to store it in the data storage module; feedback the current working status information to the central control module; the communication module is used as the communication interface between the robot system and the outside world, and can complete dual communication with the central control module. The robot communicates with the central control module, which can not only accept external control instructions, but also feedback the current working status information of the robot to the outside world; the data storage module is used to accept instructions from the central control module, feedback the current working status information to the central control module, and store large-scale data from the data acquisition module; the central control module is used as the decision-making center of the robot system, which can not only obtain the working status information of the remaining modules, but also send instructions to the remaining modules to coordinate the work of each module; the target detection model is used as a logical model for the robot to distinguish data, which can identify targets such as people or objects in the image captured by the camera, and can receive model parameter update instructions from the central control module, update the model parameters that need to be updated, and realize online update.

[0044] In this embodiment, the communication module in the robot system transmits the target detection instruction to the central control center of the robot, so that the central control center controls the target detection model in the robot system to execute the target detection instruction. When the central control center receives the target detection instruction, it will also start real-time monitoring to obtain the status of other related modules in the robot system, such as data storage capacity and data transmission efficiency, and initialize the data of each related module, such as GPIO settings and camera module configuration.

[0045] In this embodiment, the target detection model in the robot system is used to perform daily target detection tasks, and when performing the target detection task, the confidence of the target detection in the current frame is obtained, and the confidence of the target detection in the current frame is judged against a preset threshold.

[0046] Step 102: When it is determined that the confidence level of the current target detection is less than a preset threshold, the camera is controlled to collect and store image data of the current frame, and the collected image data reaching a preset amount are manually annotated.

[0047] In this embodiment, the preset threshold is set to 70%. When performing target detection, if it is determined that the confidence level of target detection in the current frame is less than the preset threshold level of 70%, the camera is controlled to collect image data of the current frame, and the image data of the current frame is stored in the data storage module; if it is determined that the confidence level of target detection in the current frame is greater than or equal to the preset threshold level of 70%, the target detection task continues to be performed without controlling the camera to collect image data of the current frame.

[0048] In this embodiment, the image data collected that reaches a preset number are manually annotated; specifically, the storage capacity of the image data with low confidence collected in the data storage module is obtained in real time, and the storage capacity of the image data with low confidence collected is compared with the preset number, wherein the preset number can be set according to demand. When the storage capacity of the image data with low confidence collected is greater than or equal to the preset number, the image data that reaches the preset number is sent to the server through the communication module, so that the data engineer manually annotates the image data through the server; when the storage capacity is less than the preset number, the camera continues to collect and store the image data with low confidence, and the current storage capacity is updated in real time until the storage capacity reaches the preset number.

[0049] As an example of this embodiment, because image data with poor confidence is often misclassified, it needs to be re-labeled. In this embodiment, the standard for classifying image data with poor confidence is that data engineers perform labeling to make the image data more representative of difficult samples, which is used to solve the classification problem of difficult samples.

[0050] Step 103: Fusing the manually annotated image data with historical image data to generate fused image data.

[0051] In this embodiment, when the robot needs to update the parameters of the target detection model, it will send the parameter update instruction to the server through the communication module, so that after the data engineer receives the parameter update instruction, the manually annotated image data and the historical image data will be CutMixed; the key part of the manually annotated image data is selected and the key part is cropped to obtain the key part data; since each image data has a label box as a label, the key part of the manually annotated image data is cropped and pasted into the historical image data. In this process, the pasted key part data must fall within the label box of the historical image data so that the key part data is fused into the same type of historical image data to generate fused image data. Among them, the selected key part is the area with the largest difference in the two pictures, and in the process of fusing the two images, the historical image data does not need to participate in the CutMix process. The historical image data is equivalent to the base data, such as Figure 3 shown.

[0052] As an example of this embodiment, CutMix fusion is to cut out part of an image and paste it on the key part of a new image, such as Figure 3 CutMix fusion in the example above selects the head of a cat image and cuts it off and pastes it on the head of a dog image. The purpose is to prevent the model from not paying attention to the most important area of ​​the model. Cutout can make the model focus on the area where the target is difficult to distinguish, and the image enhanced by CutMix can change the model's focus area on the two images. The fused area, that is, the key part, should select the most important area that can best distinguish the two images, such as Figure 3 The cat and dog in the image may have similar bodies and tails, but the details of their heads are very different. Therefore, the heads of the two images are selected for CutMix fusion to generate fused image data.

[0053] Step 104: Compare and learn the fused image data and the historical image data to obtain key-value pairs of model parameters, and update the target detection model according to the key-value pairs.

[0054] In this embodiment, the data engineer uses historical data to train the target detection model on the server, and also uses the fused image data generated by CutMix fusion to perform an additional comparative learning task. Specifically, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are of the same type is output, such as Figure 4 shown.

[0055] As an example of this embodiment, contrastive learning is to find the difference between the data to be compared and turn this difference into information for the target detection model to learn. Figure 4 Two pictures are input in , one is a bus image and the other is a truck image. The image data of the compartments of buses and trucks are relatively similar, while the images of the front of the two vehicles are quite different. Therefore, comparative learning is to find the front images of the two vehicles, so that the target detection model can focus on the front area of ​​the vehicle and output the probability that the two pictures belong to a certain category. Among them, y and y' correspond to the probabilities that the two pictures belong to a specific category, and the closer y and y' of the two pictures are, the better. This allows the target detection model to learn which areas of the same category to focus on for different examples, thereby improving the classification accuracy of difficult samples and the final target detection confidence. At the same time, without expanding the amount of data, the generalization ability of the target detection model for old task capabilities is improved, which provides ideas for the robot's completely offline comparative learning.

[0056] In this embodiment, the target detection model is trained based on contrastive learning, so that the target detection model is iteratively trained in the server to obtain a new target detection model.

[0057] In this embodiment, according to the newly obtained target detection model, its model parameters are integrated into key-value pairs, and the key-value pairs of the integrated target detection model are sent to the robot system through the communication module, so that the robot system can update the target detection model inside the robot online after obtaining the key-value pairs of the new target detection model.

[0058] In this embodiment, when transmitting model parameters, it is only necessary to send the model parameters that need to be updated to the robot system in the form of key-value pairs to realize online model update. The robot's model parameters do not need traditional repeated iterative deployment. The existing model parameters can be iteratively updated through the key-value pairs sent by the server, which simplifies the model update process.

[0059] In this embodiment, the target detection task and data collection are completed based on the existing robot system. The target is identified while data is collected. The newly collected image data and historical image data are used for CutMix data fusion and comparative learning to improve the classification accuracy of difficult samples and the robustness of the model.

[0060] See also Figure 2 , Figure 2 is a structural diagram of an embodiment of a device for updating a target detection model provided by the present invention, such as Figure 2As shown, the device includes a confidence acquisition module 201, an image data acquisition module 202, an image fusion module 203 and a comparison module 204, which are specifically as follows:

[0061] The confidence acquisition module 201 is used to perform target detection through the target detection model, obtain the confidence of the current target detection, and compare the confidence of the current target detection with a preset threshold;

[0062] The image data acquisition module 202 is used to control the camera to acquire and store the image data of the current frame when it is determined that the confidence of the current target detection is less than a preset threshold, and to manually annotate the acquired image data reaching a preset amount;

[0063] An image fusion module 203 is used to fuse the manually annotated image data with historical image data to generate fused image data;

[0064] The comparison module 204 is used to compare and learn the fused image data with the historical image data to obtain key-value pairs of model parameters, and update the target detection model according to the key-value pairs.

[0065] In this embodiment, the image data acquisition module 202 is used to manually annotate the image data collected that reaches a preset amount. Specifically, the storage capacity of the image data is obtained in real time, and the storage capacity is compared with the preset amount; when the storage capacity is greater than or equal to the preset amount, the image data that reaches the preset amount is sent to the server side so that the image data is manually annotated; when the storage capacity is less than the preset amount, wait for the next storage and update the current storage capacity in real time.

[0066] In this embodiment, the image fusion module 203 is used to fuse the manually annotated image data with the historical image data to generate fused image data; specifically, the key part of the manually annotated image data is selected, and the key part is cropped to obtain the key part data; the key part data is fused into the historical image data of the same type to generate fused image data.

[0067] In this embodiment, the comparison module 204 is used to compare and learn the fused image data with the historical image data; specifically, when the target detection model is trained, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are the same type of data is output.

[0068] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0069] It should be noted that the above-mentioned embodiment of the updating device of the target detection model is only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0070] Based on the above-mentioned embodiment of the target detection model updating method, another embodiment of the present invention provides a target detection model updating terminal device, the target detection model updating terminal device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, when the processor executes the computer program, it implements the target detection model updating method of any embodiment of the present invention.

[0071] Exemplarily, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the update terminal device of the target detection model.

[0072] The target detection model update terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The target detection model update terminal device may include, but is not limited to, a processor and a memory.

[0073] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the update terminal device of the target detection model, and uses various interfaces and lines to connect various parts of the update terminal device of the entire target detection model.

[0074] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the target detection model update terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0075] Based on the above-mentioned embodiment of the target detection model updating method, another embodiment of the present invention provides a storage medium, wherein the storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the target detection model updating method of any embodiment of the present invention.

[0076] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0077] In summary, the present invention provides a method, device, equipment and storage medium for updating a target detection model. The target detection model is used to perform target detection, obtain the confidence of the current target detection, and compare the confidence of the current target detection with a preset threshold; when the confidence of the current target detection is judged to be less than the preset threshold, the camera is controlled to collect and store the image data of the current frame, and the collected image data reaching a preset number are manually annotated; the manually annotated image data is fused with the historical image data to generate fused image data; the fused image data is compared and studied with the historical image data to obtain the key-value pairs of the model parameters, and the target detection model is updated according to the key-value pairs. Compared with the prior art, the present invention improves the robustness of the model by fusing and comparing the collected image data with the historical image data, so that the classification of difficult samples is more accurate.

[0078] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention. These improvements and substitutions should also be regarded as the scope of protection of the present invention.

Claims

1. A method for updating a target detection model, characterized in that: include: Performing target detection through a target detection model, obtaining the confidence of the current target detection, and comparing the confidence of the current target detection with a preset threshold; When it is determined that the confidence level of the current target detection is less than a preset threshold, the camera is controlled to collect and store image data of the current frame, and the collected image data reaching a preset amount are manually annotated; The manually annotated image data is fused with the historical image data to generate fused image data; wherein a key part of the manually annotated image data is selected and the key part is cropped to obtain key part data; the key part data is fused with the historical image data of the same type to generate fused image data; The fused image data and the historical image data are compared and learned to obtain key-value pairs of model parameters, and the target detection model is updated according to the key-value pairs; wherein, when the target detection model is trained, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are of the same type of data is output; the target detection model is trained based on comparative learning, so that the target detection model is iteratively trained in the server to obtain a new target detection model; according to the newly obtained target detection model, its model parameters are integrated into key-value pairs.

2. A method for updating a target detection model as claimed in claim 1, characterized in that: Manually labeling the collected image data that reaches a preset amount is specifically: Acquiring the storage amount of the image data in real time, comparing the storage amount with a preset amount, and when the storage amount is greater than or equal to the preset amount, sending the image data reaching the preset amount to a server side so as to manually annotate the image data; When the storage amount is less than the preset amount, wait for the next storage and update the current storage amount in real time.

3. A device for updating a target detection model, characterized in that: include: Confidence acquisition module, image data acquisition module, image fusion module and comparison module; The confidence acquisition module is used to perform target detection through the target detection model, obtain the confidence of the current target detection, and compare the confidence of the current target detection with a preset threshold; The image data acquisition module is used to control the camera to acquire and store the image data of the current frame when it is determined that the confidence level of the current target detection is less than a preset threshold, and to manually annotate the acquired image data reaching a preset amount; The image fusion module is used to fuse the manually annotated image data with the historical image data to generate fused image data; wherein, a key part of the manually annotated image data is selected and the key part is cropped to obtain key part data; and the key part data is fused into the historical image data of the same type to generate fused image data; The comparison module is used to compare and learn the fused image data with the historical image data to obtain key-value pairs of model parameters, and update the target detection model according to the key-value pairs; wherein, when the target detection model is trained, the fused image data and the historical image data are input into the target detection model for comparison, and the probability that the fused image data and the historical image data are of the same type of data is output; the target detection model is trained based on comparative learning so that the target detection model is iteratively trained in the server to obtain a new target detection model; according to the newly obtained target detection model, its model parameters are integrated into key-value pairs.

4. The updating device for a target detection model according to claim 3, characterized in that: The image data acquisition module is used to manually annotate the image data that has been acquired and reaches a preset amount, specifically: Acquiring the storage amount of the image data in real time, comparing the storage amount with a preset amount, and when the storage amount is greater than or equal to the preset amount, sending the image data reaching the preset amount to a server side so as to manually annotate the image data; When the storage amount is less than the preset amount, wait for the next storage and update the current storage amount in real time.

5. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for updating the target detection model as described in any one of claims 1 to 2 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the target detection model updating method according to any one of claims 1 to 2.

Images