Machine learning-based desktop non-learning behavior identification method and device
By adopting machine learning-based recognition methods in the laboratory information management system, using the YOLO-v5 picture classification model to judge non-learning behaviors on students' desktops, the problem of inefficient identification and management in the existing technology is solved, efficient non-learning behavior recognition and intervention is achieved, and the quality of laboratory teaching is improved.
Patent Information
- Application Number
- CN202411761225.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-03-28
AI Technical Summary
In the prior art, middle school students cannot efficiently identify and manage non-learning behaviors of computer desktops during the laboratory operation, which affects the quality of laboratory teaching and students' learning effects.
Using machine learning-based recognition method, the screenshot is taken through the client, and the screenshot picture is entered into the YOLO-v5 picture classification model, the probability that the picture is a game picture, and the picture with a probability greater than or equal to the preset threshold is uploaded to the server as a game picture. The server sends reminder information about non-learning behavior to the client and the supervision and spot check interface.
It realizes efficient and timely identification and management of students' non-learning behaviors, and improves the quality of laboratory teaching and students' learning effects.
Smart Images
Figure CN119723167A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of recognition technology, and in particular to a method and device for recognizing desktop non-learning behaviors based on machine learning. Background Art
[0002] With the rapid development of educational informatization, laboratory teaching has become an important part of university education. In order to improve the quality of laboratory teaching and management efficiency, many schools have adopted the laboratory information management system LISS. The class inspection and supervision function in the LISS system can project the computer system desktop of the laboratory to the school's large screen in real time, providing teachers or administrators with a real-time entrance to view and supervise students' computer use, which can improve the transparency and management level of laboratory teaching to a certain extent.
[0003] However, in actual use, it was found that some students engaged in non-learning behaviors such as playing games and browsing irrelevant web pages while using the computer in the laboratory. These behaviors seriously affected the quality of laboratory teaching and students' learning outcomes. Although teachers and administrators can discover and manage these problems through the class inspection and supervision function, the identification and management efficiency is low and there is a lack of efficient identification and management solutions.
[0004] Therefore, how to efficiently identify non-learning behaviors on the desktop while students are using the computer is a technical problem that needs to be solved urgently. Summary of the invention
[0005] The present application provides a method and device for identifying desktop non-learning behaviors based on machine learning, which can solve the technical problem in the prior art that students cannot efficiently identify and manage computer desktop non-learning behaviors during computer use.
[0006] In a first aspect, an embodiment of the present application provides a method for identifying desktop non-learning behaviors based on machine learning, which is used on a client. The method for identifying desktop non-learning behaviors based on machine learning includes:
[0007] The client takes a screenshot of the screen to obtain a window image;
[0008] Input the window image to the preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen;
[0009] Upload the window image whose game screen probability is greater than or equal to the preset probability threshold as the game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface;
[0010] A non-learning behavior reminder pop-up window is displayed on the screen according to the reminder information sent by the server.
[0011] In combination with the first aspect, in one implementation, a window image whose game screen probability is greater than or equal to a preset probability threshold is uploaded to a server as a game image, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface, and further includes:
[0012] When a game image is first recognized, a game start mark is marked on the first recognized game image, and the image is uploaded to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface;
[0013] After the game image is recognized for the first time, every time the game image is recognized for M consecutive times, the in-game logo is marked in the game image recognized for the Mth time and uploaded to the server.
[0014] In one embodiment, the method further comprises:
[0015] If a non-game picture is identified after a game picture is identified whose game screen probability is less than a preset probability threshold, a game end mark is marked in a game picture before the non-game picture and uploaded to the server, so that the server stops sending non-learning behavior reminder information to the client and the supervision and spot check interface.
[0016] In one embodiment, before inputting the window image to the preset YOLO-v5 image classification model, it also includes modifying and packaging the YOLO-v5 image classification model, specifically including:
[0017] Modify the loss function of the YOLO-v5 image classification model to EloU;
[0018] Modify the training script hyperparameters of the YOLO-v5 image classification model, wherein the training script hyperparameters include batch size batch-size, image size img-size, learning rate lr, and number of iterations epochs;
[0019] Training the YOLO-v5 image classification model according to a preset classification training data set;
[0020] When the recall rate of the YOLO-v5 image classification model meets the preset calibration value, the weight of the YOLO-v5 image classification model is obtained, and the YOLO-v5 image classification model is packaged into an exe program.
[0021] In one implementation, the client takes a screenshot of the screen to obtain a window image, including:
[0022] The client takes screenshots of active windows and full-screen windows on the screen at preset time intervals to obtain corresponding window images.
[0023] In a second aspect, an embodiment of the present application provides a method for identifying desktop non-learning behaviors based on machine learning, which is used for a server. The method for identifying desktop non-learning behaviors based on machine learning includes:
[0024] Receive game pictures sent by the client;
[0025] According to the game picture, reminder information of non-learning behavior is sent to the corresponding client and the supervisory spot check interface, so that the corresponding client displays a non-learning behavior reminder pop-up window on the screen, and the supervisory spot check interface displays warning information on the desktop of the client corresponding to the game picture.
[0026] In conjunction with the second aspect, in one implementation, the method further includes:
[0027] When receiving a game picture marked with a game end mark sent by the client, stop sending reminder information of non-learning behavior to the corresponding client and the supervision spot check interface.
[0028] In one embodiment, after receiving the game image sent by the client, the method further includes:
[0029] The game images and the probability that they are game images are saved as training data for training the YOLO-v5 image classification model.
[0030] In a third aspect, an embodiment of the present application provides a method for identifying desktop non-learning behaviors based on machine learning, which is used for supervising a spot check interface. The method for identifying desktop non-learning behaviors based on machine learning includes:
[0031] Receive non-learning line reminder information sent by the server based on the game image of the client;
[0032] According to the reminder information, a warning message is displayed on the client desktop corresponding to the game picture.
[0033] In a fourth aspect, an embodiment of the present application provides a device for identifying desktop non-learning behaviors based on machine learning, which is used in a client. The device for identifying desktop non-learning behaviors based on machine learning includes:
[0034] The screenshot module is used by the client to take a screenshot of the screen and obtain a window image;
[0035] The recognition module is used to input the window image into the preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen;
[0036] A sending module, which is used to upload the window image whose game screen probability is greater than or equal to a preset probability threshold as a game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface;
[0037] A reminder module is used to display a non-learning behavior reminder pop-up window on the screen according to the reminder information sent by the server.
[0038] The embodiment of the present application provides a method and device for identifying desktop non-learning behavior based on machine learning, which comprises taking a screenshot of the screen through a client to obtain a window image; inputting the window image into a preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen; taking the window image with a probability of being a game screen greater than or equal to a preset probability threshold as a game picture, uploading it to a server, so that the server sends a reminder message of non-learning behavior to the client and a supervisory spot check interface; displaying a non-learning behavior reminder pop-up window on the screen according to the reminder message sent by the server, thereby realizing efficient and early discovery of non-learning behavior in the client, and intervening as soon as possible, thereby improving the quality of laboratory teaching and students' learning effects. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 This is a flow chart of an embodiment of a method for identifying desktop non-learning behaviors based on machine learning implemented by a client in the present application;
[0040] Figure 2 This is a diagram of the training process of the YOLO-v5 image classification model for this application;
[0041] Figure 3 This is a flow chart of an embodiment of a method for identifying desktop non-learning behaviors based on machine learning implemented by a server in the present application;
[0042] Figure 4 A flowchart of an embodiment of a method for identifying desktop non-learning behaviors based on machine learning implemented through a supervision and spot check interface in this application;
[0043] Figure 5 This application is a device for identifying desktop non-learning behaviors based on machine learning used on the client. DETAILED DESCRIPTION
[0044] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0045] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.
[0046] In a first aspect, an embodiment of the present application provides a method for identifying desktop non-learning behaviors based on machine learning, which can be executed by a client.
[0047] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the method for identifying desktop non-learning behavior based on machine learning in this application. Figure 1 As shown, the recognition method of desktop non-learning behavior based on machine learning includes:
[0048] Step S101: The client takes a screenshot of the screen to obtain a window image.
[0049] In one embodiment, the client periodically takes screenshots of active windows and full-screen windows on its own screen at preset time intervals to obtain corresponding window images.
[0050] Step S102: input the window image into a preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen.
[0051] Exemplarily, the probability that the window image output by the YOLO-v5 image classification model is a game screen can be marked in the upper left corner of the window image.
[0052] It is worth noting that before inputting the window image into the preset YOLO-v5 image classification model, it also includes modifying and packaging the YOLO-v5 image classification model, specifically including:
[0053] like Figure 2 As shown, a classification training data set of a YOLO-v5 image classification model is prepared in advance, the loss function of the YOLO-v5 image classification model is modified to EloU, and the training script hyperparameters of the YOLO-v5 image classification model are modified, wherein the training script hyperparameters include batch size batch-size, image size img-size, learning rate lr, and number of iterations epochs. Such modifications can improve the feature differentiation ability of the YOLO-v5 image classification model, thereby helping to improve the accuracy of classification.
[0054] The YOLO-v5 image classification model is trained according to a preset classification training data set; when the recall rate of the YOLO-v5 image classification model meets the preset calibration value, the weight of the YOLO-v5 image classification model is obtained, and the YOLO-v5 image classification model is packaged into an exe program.
[0055] Exemplarily, in this embodiment, batch-size, img-size, lr, and epochs are adjusted when training the YOLO-v5 image classification model to obtain the optimal weight file based on the YOLO-v5 image classification model, achieving an accuracy of 85% for web games based on screenshot window images, 100% for windows built-in card games, and 70% for online games. At the same time, the memory occupied during operation is only about 300M, and the single recognition time is less than 2s after the model is loaded.
[0056] Step S103: Upload the window image whose game screen probability is greater than or equal to the preset probability threshold as the game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface.
[0057] Specifically, step S103 specifically includes the following steps: when the game image is recognized for the first time, marking the game start logo on the first recognized game image, and uploading it to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface; after the game image is recognized for the first time, every M consecutive times the game image is recognized, marking the game logo in the Mth recognized game image, and uploading it to the server.
[0058] Exemplarily, in this embodiment, the probability threshold is set to 60%, and the value of M can be 10. When the client first recognizes that the probability of the window image being a game screen is greater than or equal to 60%, the window image will be used as a game image and marked with a game start logo, and the game image and the corresponding recognition result will be uploaded to the server. The client will subsequently recognize the window image once every half a minute, and every 10 consecutive times it recognizes a game image with a probability of the game screen greater than or equal to the preset probability threshold, the game image recognized for the 10th time will be marked with an in-game logo, and the game image and the corresponding recognition result will be uploaded to the server. The probability of each window image being a game screen is marked in the window image, so when the game image is sent to the server, the probability of the game image being a game screen is also sent to the server as a recognition result.
[0059] It is worth mentioning that the laboratory information management LISS system can be used on multiple clients. The supervision and spot check interface includes the desktops of each client being supervised and checked. When any client detects non-learning behavior for the first time, the server will send a reminder message of the non-learning behavior to the corresponding client and the supervision and spot check interface at the same time.
[0060] Step S104: The client displays a non-learning behavior reminder pop-up window on the screen according to the reminder information sent by the server, thereby reminding the user of the client to stop the game behavior.
[0061] It is worth mentioning that after the supervision and spot check interface receives the reminder information of non-learning behavior sent by the server, it will display a red warning message on the client desktop corresponding to the game picture according to the reminder information, thereby reminding teachers or administrators of the clients with gaming behavior to facilitate management by teachers or administrators.
[0062] In one embodiment, the method also includes: if a non-game picture with a game screen probability less than a preset probability threshold is identified after the game picture is identified, indicating that the game behavior on the client has stopped, then marking a game end mark in a game picture before the non-game picture and uploading it to the server, so that the server stops sending reminder information of non-learning behavior to the client and the supervision and spot check interface.
[0063] It is worth mentioning that the server will save the game images uploaded by the client for a certain period of time, and can subsequently export them to retrain and update the YOLO-v5 image classification model. When deploying, you only need to update the weight file in the installation package, thereby improving the accuracy of the YOLO-v5 image classification model in recognizing game images.
[0064] The method for identifying desktop non-learning behaviors based on machine learning provided in the embodiment of the present application can efficiently and timely discover students' non-learning behaviors by identifying the client window screen through the YOLO-v5 image classification model, and can intervene in students' non-learning behaviors by displaying a non-learning behavior reminder pop-up window on the client, thereby improving the quality of laboratory teaching and students' learning effects. The YOLO-v5 image classification model is a lightweight AI model and can only run in the CPU, which puts little burden on computing resources. The YOLO-v5 image classification model has learning ability and supports continuous learning to improve recognition accuracy during use. When training the AI model, batch-size, img-size, lr, and epochs are adjusted to obtain the best weight file based on YOLO-v5s, thereby improving the recognition accuracy of the YOLO-v5 image classification model.
[0065] In a second aspect, an embodiment of the present application also provides a method for identifying desktop non-learning behaviors based on machine learning, which is used for a server.
[0066] In one embodiment, referring to Figure 3 , Figure 3 This is the process intention of an embodiment of the method for identifying desktop non-learning behavior based on machine learning in this application. Figure 3 As shown, the recognition method of desktop non-learning behavior based on machine learning includes:
[0067] Step S201, receiving a game picture sent by a client.
[0068] Step S202: Send non-learning behavior reminder information to the corresponding client and the supervisory spot check interface according to the game picture, so that the corresponding client displays a non-learning behavior reminder pop-up window on the screen, and the supervisory spot check interface displays warning information on the desktop of the client corresponding to the game picture.
[0069] Furthermore, in one embodiment, the method further includes:
[0070] When receiving a game picture marked with a game end mark sent by the client, stop sending reminder information of non-learning behavior to the corresponding client and the supervision spot check interface.
[0071] Furthermore, in one embodiment, after receiving the game picture sent by the client, the method further includes:
[0072] The game images and the probability that they are game images are saved as training data for training the YOLO-v5 image classification model.
[0073] On the third aspect, the embodiment of the present application also provides a method for identifying desktop non-learning behaviors based on machine learning, which is used for supervising the spot check interface.
[0074] In one embodiment, referring to Figure 4 , Figure 4 This is the process intention of an embodiment of the method for identifying desktop non-learning behavior based on machine learning in this application. Figure 4 As shown, the recognition method of desktop non-learning behavior based on machine learning includes:
[0075] Step S301: receiving non-learning line reminder information sent by the server based on the game picture of the client.
[0076] Step S302: display warning information on the client desktop corresponding to the game image according to the reminder information.
[0077] In a fourth aspect, an embodiment of the present application also provides a device for identifying desktop non-learning behaviors based on machine learning, which is used on a client.
[0078] In one embodiment, referring to Figure 5 , Figure 5 This is a schematic block diagram of the structure of an embodiment of a device for identifying desktop non-learning behaviors based on machine learning in this application. Figure 5 As shown, the device for identifying desktop non-learning behavior based on machine learning includes:
[0079] The screenshot module is used by the client to take a screenshot of the screen and obtain a window image;
[0080] The recognition module is used to input the window image into the preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen;
[0081] A sending module, which is used to upload the window image whose game screen probability is greater than or equal to a preset probability threshold as a game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface;
[0082] A reminder module is used to display a non-learning behavior reminder pop-up window on the screen according to the reminder information sent by the server.
[0083] In one embodiment, the sending module is further used for:
[0084] When a game image is first recognized, a game start mark is marked on the first recognized game image, and the image is uploaded to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface;
[0085] After the game image is recognized for the first time, every time the game image is recognized for M consecutive times, the in-game logo is marked in the game image recognized for the Mth time and uploaded to the server.
[0086] In one embodiment, the sending module is further used for:
[0087] If a non-game picture is identified after a game picture is identified whose game screen probability is less than a preset probability threshold, a game end mark is marked in a game picture before the non-game picture and uploaded to the server, so that the server stops sending non-learning behavior reminder information to the client and the supervision and spot check interface.
[0088] In one embodiment, the device is also used for:
[0089] Modify the loss function of the YOLO-v5 image classification model to EloU;
[0090] Modify the training script hyperparameters of the YOLO-v5 image classification model, wherein the training script hyperparameters include batch size batch-size, image size img-size, learning rate lr, and number of iterations epochs;
[0091] Training the YOLO-v5 image classification model according to a preset classification training data set;
[0092] When the recall rate of the YOLO-v5 image classification model meets the preset calibration value, the weight of the YOLO-v5 image classification model is obtained, and the YOLO-v5 image classification model is packaged into an exe program.
[0093] In one embodiment, the screenshot module is further used for:
[0094] The client takes screenshots of active windows and full-screen windows on the screen at preset time intervals to obtain corresponding window images.
[0095] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.
[0096] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.
[0097] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.
[0098] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; the “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.
[0099] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.
[0100] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present application.
[0101] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A method for identifying desktop non-learning behavior based on machine learning, characterized in that: The method is used on the client, and the method for identifying desktop non-learning behavior based on machine learning includes: The client takes a screenshot of the screen to obtain a window image; Input the window image to the preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen; Upload the window image whose game screen probability is greater than or equal to the preset probability threshold as the game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface; A non-learning behavior reminder pop-up window is displayed on the screen according to the reminder information sent by the server.
2. The method for identifying desktop non-learning behaviors based on machine learning as claimed in claim 1, characterized in that: The window image whose game screen probability is greater than or equal to a preset probability threshold is used as a game image and uploaded to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface, and also includes: When a game image is first recognized, a game start mark is marked on the first recognized game image, and the image is uploaded to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervisor spot check interface; After the game image is recognized for the first time, every time the game image is recognized for M consecutive times, the in-game logo is marked in the game image recognized for the Mth time and uploaded to the server.
3. The method for identifying desktop non-learning behavior based on machine learning as claimed in claim 2, characterized in that: The method further includes: If a non-game picture is identified after a game picture is identified whose game screen probability is less than a preset probability threshold, a game end mark is marked in a game picture before the non-game picture and uploaded to the server, so that the server stops sending non-learning behavior reminder information to the client and the supervision and spot check interface.
4. The method for identifying desktop non-learning behavior based on machine learning as claimed in claim 1, characterized in that: Before inputting the window image to the preset YOLO-v5 image classification model, the YOLO-v5 image classification model is modified and packaged, specifically including: Modify the loss function of the YOLO-v5 image classification model to EloU; Modify the training script hyperparameters of the YOLO-v5 image classification model, wherein the training script hyperparameters include batch size batch-size, image size img-size, learning rate lr, and number of iterations epochs; Training the YOLO-v5 image classification model according to a preset classification training data set; When the recall rate of the YOLO-v5 image classification model meets the preset calibration value, the weight of the YOLO-v5 image classification model is obtained, and the YOLO-v5 image classification model is packaged into an exe program.
5. The method for identifying desktop non-learning behaviors based on machine learning as claimed in claim 1, characterized in that: The client takes a screenshot of the screen to obtain a window image, including: The client takes screenshots of active windows and full-screen windows on the screen at preset time intervals to obtain corresponding window images.
6. A method for identifying desktop non-learning behavior based on machine learning, characterized in that: The method is used for a server, and the method for identifying desktop non-learning behavior based on machine learning includes: Receive game pictures sent by the client; According to the game picture, reminder information of non-learning behavior is sent to the corresponding client and the supervisory spot check interface, so that the corresponding client displays a non-learning behavior reminder pop-up window on the screen, and the supervisory spot check interface displays warning information on the desktop of the client corresponding to the game picture.
7. The method for identifying desktop non-learning behaviors based on machine learning as claimed in claim 8, characterized in that: The method further includes: When receiving a game picture marked with a game end mark sent by the client, stop sending reminder information of non-learning behavior to the corresponding client and the supervision spot check interface.
8. The method for identifying desktop non-learning behaviors based on machine learning as claimed in claim 6, characterized in that: After receiving the game picture sent by the client, it also includes: The game images and the probability that they are game images are saved as training data for training the YOLO-v5 image classification model.
9. A method for identifying desktop non-learning behavior based on machine learning, characterized in that: The method is used for supervising the spot check interface, and the method for identifying desktop non-learning behavior based on machine learning includes: Receive non-learning line reminder information sent by the server based on the game image of the client; According to the reminder information, a warning message is displayed on the client desktop corresponding to the game picture.
10. A desktop non-learning behavior recognition device based on machine learning, characterized in that: The device is used in the client, and the desktop non-learning behavior recognition device based on machine learning includes: The screenshot module is used by the client to take a screenshot of the screen and obtain a window image; The recognition module is used to input the window image into the preset YOLO-v5 image classification model to obtain the probability that the window image is a game screen; A sending module, which is used to upload the window image whose game screen probability is greater than or equal to a preset probability threshold as a game image to the server, so that the server sends a reminder message of non-learning behavior to the client and the supervision and spot check interface; A reminder module is used to display a non-learning behavior reminder pop-up window on the screen according to the reminder information sent by the server.