Training method and device for definition of inspection picture based on AI (Artificial Intelligence)

Abstract

The invention provides an AI-based inspection picture definition training method and device. The method comprises the following steps: taking inspection pictures of operation and maintenance machine room equipment as a first sample data set; dividing the first sample data set into a first training data set and a first test data set according to a preset proportion; performing adversarial training on the generator and the discriminator through the first training data set by utilizing the principle of a generative adversarial network model to obtain a machine room equipment photo definition improvement model; inputting the first test data set into a machine room equipment photo definition improvement model to obtain a high-definition picture; calculating a quantitative index PSNR according to the high-definition picture and the original high-definition picture; judging whether the quantitative index PSNR is in a preset index range or not; if yes, the machine room equipment photo definition improvement model meets the requirement; the definition and resolution of a fuzzy and low-resolution inspection picture of the electronic operation and maintenance machine room equipment are improved, and a more reliable data basis is provided for accurate recording and analysis of the state of the machine room equipment.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an AI-based inspection picture clarity training method and device. Background Art

[0002] With the rapid development of information technology, electronic operation and maintenance rooms, as the core infrastructure of modern enterprises, play a vital role. During the operation and maintenance inspection of the room, equipment photos are taken regularly to record and monitor the equipment status.

[0003] In the current electronic operation and maintenance room inspection, due to the constraints of various factors such as environmental conditions, operating techniques and equipment conditions, the clarity and resolution of photos are often not high. These blurry and low-resolution photos may cause difficulties and misunderstandings in the subsequent analysis, reporting and troubleshooting process.

[0004] The image clarity enhancement tools currently on the market have a poor effect on improving blurry photos taken during computer room inspections due to the lack of big data sample training for computer room equipment. There are also problems such as low improvement success rate and confusion in the details of the improved images. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an AI-based inspection image clarity training method and device, which can improve the clarity and resolution of blurry, low-resolution inspection images of electronic operation and maintenance room equipment, thereby providing a more reliable data basis for accurate recording and subsequent analysis of the status of equipment in the room.

[0006] In a first aspect, an embodiment of the present invention provides an AI-based inspection picture clarity training method, the method comprising:

[0007] Collect inspection pictures of equipment in the operation and maintenance room and use them as the first sample data set;

[0008] Dividing the first sample data set into a first training data set and a first test data set according to a preset ratio;

[0009] By utilizing the principle of the generative adversarial network model, the generator and the discriminator are adversarially trained using the first training data set to obtain a model for improving the clarity of photos of equipment in a computer room.

[0010] Furthermore, by using the principle of the generative adversarial network model, the generator and the discriminator are adversarially trained using the first training data set to obtain a model for improving the clarity of photos of equipment in a computer room, including:

[0011] Performing image resolution degradation processing on the original inspection images in the first training data set to obtain low-resolution images corresponding to the original inspection images;

[0012] Inputting the low-definition picture into the generator to generate a high-definition picture;

[0013] Inputting the high-definition picture and the original inspection picture into the discriminator, and outputting an evaluation result, wherein the evaluation result is used to evaluate the authenticity of the generated high-definition picture;

[0014] The generator and the discriminator are iteratively optimized through adversarial learning, and after training, the clarity improvement model of the equipment room photos is obtained.

[0015] Furthermore, before dividing the first sample data set into a first training data set and a first test data set according to a preset ratio, the method further includes:

[0016] The first sample data set is preprocessed, and the preprocessing includes resizing, normalization, and cropping to ensure that the size of the original inspection images in the first training data set remains consistent.

[0017] Furthermore, after performing adversarial training on the generator and the discriminator through the first training data set to obtain a model for improving the clarity of photos of equipment in a computer room, the method further includes:

[0018] Inputting the first test data set into the computer room equipment photo clarity improvement model to obtain a high-definition picture;

[0019] Calculate the quantitative indicator PSNR based on the high-definition picture and the original high-definition picture;

[0020] Determine whether the quantitative indicator PSNR is within a preset indicator range;

[0021] If yes, it is determined that the model for improving the clarity of the equipment room photos meets the requirements;

[0022] If not, it is determined that the computer room equipment photo clarity improvement model does not meet the requirements.

[0023] Further, a quantitative indicator PSNR is calculated based on the high-definition picture and the original high-definition picture, including:

[0024] Calculating a mean square error based on the high-definition image and the original high-definition image;

[0025] The quantitative indicator PSNR is calculated according to the maximum value of the mean square error and the image pixel value.

[0026] Further, after determining that the model for improving the clarity of the equipment room photos does not meet the requirements, the method further includes:

[0027] Adding multiple inspection pictures to the first sample data set to obtain a second sample data set;

[0028] Dividing the second sample data set into a second training data set and a second test data set according to a preset ratio;

[0029] Using the principle of the generative adversarial network model, the generator and the discriminator are re-trained adversarially using the second training data set to obtain a retrained model for improving the clarity of photos of equipment in the computer room;

[0030] After inputting the second test data set into the retrained computer room equipment photo clarity improvement model, and calculating the quantitative indicator PSNR, until the quantitative indicator PSNR is within the preset indicator range, the retrained computer room equipment photo clarity improvement model meets the requirements.

[0031] Further, if yes, after determining that the computer room equipment photo definition improvement model meets the requirements, the method further includes:

[0032] Get the inspection pictures to be tested;

[0033] Input the inspection picture to be detected into the equipment room equipment photo clarity improvement model to obtain a high-definition inspection picture to be detected;

[0034] Get high-definition historical inspection pictures;

[0035] The state of the monitoring device is determined by comparing the high-definition inspection picture to be detected with the high-definition historical inspection picture, and the high-definition inspection picture to be detected is stored as the high-definition historical inspection picture.

[0036] In a second aspect, an embodiment of the present invention provides an AI-based inspection picture clarity training device, the device comprising:

[0037] The first acquisition module is used to collect inspection pictures of equipment in the operation and maintenance room and use them as the first sample data set;

[0038] A first division module, used to divide the first sample data set into a first training data set and a first test data set according to a preset ratio;

[0039] The first adversarial training module is used to utilize the principle of the generative adversarial network model to perform adversarial training on the generator and the discriminator through the first training data set to obtain a model for improving the clarity of photos of equipment in the computer room.

[0040] In a third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and when the processor executes the computer program, the AI-based inspection image clarity training method as described above is implemented.

[0041] In a fourth aspect, an embodiment of the present invention provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute the AI-based inspection image clarity training method as described above.

[0042] The embodiment of the present invention provides an AI-based inspection picture clarity training method and device, including: collecting inspection pictures of operation and maintenance room equipment and using them as a first sample data set; dividing the first sample data set into a first training data set and a first test data set according to a preset ratio; introducing artificial intelligence AI technology into the field of electronic operation and maintenance room equipment inspection, using the principle of a generative adversarial network model, and conducting adversarial training on a generator and a discriminator through a first training data set to obtain a picture clarity improvement model for room equipment; in an evaluation method, inputting the first test data set into the picture clarity improvement model for room equipment to obtain a high-definition picture; calculating a quantitative indicator PSNR based on the high-definition picture and the original high-definition picture; judging whether the quantitative indicator PSNR is within a preset indicator range; if so, the picture clarity improvement model for room equipment meets the requirements; if not, the picture clarity improvement model for room equipment does not meet the requirements; the clarity and resolution of blurry and low-resolution inspection pictures of electronic operation and maintenance room equipment can be improved, providing a more reliable data basis for accurate recording and subsequent analysis of the room equipment status.

[0043] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A flow chart of a method for training the clarity of inspection images based on AI provided in Embodiment 1 of the present invention;

[0047] Figure 2 A flow chart of a method for evaluating the clarity of inspection images provided in Embodiment 1 of the present invention;

[0048] Figure 3 Schematic diagram of the AI-based inspection image clarity training device provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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.

[0050] The image clarity enhancement tools currently on the market have a poor effect on the enhancement of blurred photos of computer room inspections due to the lack of big data sample training for computer room equipment. There are problems such as low enhancement success rate and confusion of image details after enhancement. In order to solve this problem, the present invention applies to introduce artificial intelligence AI technology into the field of electronic operation and maintenance computer room equipment inspection, and based on the high-definition image generation technology of the generative adversarial network model, the blurred and low-resolution electronic operation and maintenance computer room equipment inspection images are enhanced in clarity and resolution, providing a more reliable data basis for the accurate recording and subsequent analysis of the status of computer room equipment. This application can effectively improve the clarity of more than 90% of the blurred inspection photos. The deep learning model analysis integrates the characteristics of various equipment in the computer room, and has the characteristics of fast computing speed, short training time, and the enhanced details are real and accurate.

[0051] To facilitate understanding of this embodiment, the embodiment of the present invention is described in detail below.

[0052] Embodiment 1:

[0053] Figure 1 A flow chart of the AI-based inspection image clarity training method provided in Example 1 of the present invention.

[0054] Reference Figure 1 , the method comprises the following steps:

[0055] Step S101, collecting inspection pictures of equipment in the operation and maintenance room and using them as a first sample data set;

[0056] Step S102, dividing the first sample data set into a first training data set and a first test data set according to a preset ratio;

[0057] Specifically, the equipment in the operation and maintenance room includes at least equipment servers, switches, routers, and storage devices, etc. The preset ratio may be 6:4.

[0058] The first sample data set is divided into a training data set and a test data set in a ratio of 6:4. The ratio of allocating the training data set and the test data set depends on the total amount of available data and the complexity of the task. The total amount of data and the complexity of the task involved in this application are both medium, which is suitable for 60% training data set and 40% test data set.

[0059] In addition, before step S102, the first sample data set needs to be preprocessed, and the preprocessing includes resizing, normalization, and cropping to ensure that the size of the original inspection images in the first training data set remains consistent.

[0060] The purpose of normalization is to convert the input image data into a standardized form so that the neural network can learn the features of the image more easily and help the stability and convergence speed of the training process. Specifically, it includes:

[0061] 1) Pixel value scaling: Scale the pixel value to the range of [-1, 1] and use the method of pixel value = (pixel value / 127.5)-1.0 to calculate, which matches the input shape of the equipment room photo clarity improvement model.

[0062] 2) Mean subtraction: The pixel mean is calculated on each channel and then subtracted from each pixel channel of the image to eliminate brightness and contrast differences in the image.

[0063] Step S103, using the principle of the generative adversarial network model, the generator and the discriminator are adversarially trained using the first training data set to obtain a model for improving the clarity of photos of equipment in the computer room.

[0064] Among them, the adversarial network model belongs to artificial intelligence AI technology.

[0065] In order to solve the problem of blurred and low-resolution inspection photos of equipment in the operation and maintenance computer room, this application makes full use of artificial intelligence AI technology and the principles of generative adversarial network models, and conducts adversarial training on the generator and discriminator through the first training data set to obtain a computer room equipment photo clarity improvement model, thereby realizing the conversion of blurred inspection photos into high-quality, higher-definition images, so as to improve data quality and the accuracy of subsequent analysis.

[0066] Further, step S103 includes the following steps:

[0067] Step S201, performing image resolution degradation processing on the original inspection image in the first training data set to obtain a low-definition image corresponding to the original inspection image;

[0068] Here, firstly, the original inspection picture is subjected to image resolution degradation processing, that is, the image resolution is reduced to one quarter of the original image, so as to obtain a low-definition image corresponding to the original inspection picture.

[0069] Step S202, inputting the low-definition picture into the generator to generate a high-definition picture;

[0070] Step S203, inputting the high-definition picture and the original inspection picture into the discriminator, outputting an evaluation result, and the evaluation result is used to evaluate the authenticity of the generated high-definition picture;

[0071] Specifically, the discriminator includes a convolutional neural network and an activation function, and the output evaluation result is a result between 0 and 1; when the evaluation result is close to 1, it represents true; when the evaluation result is close to 0, it represents false. For example, when the evaluation result is 0.1, it means that the image is false; when the evaluation result is 1, it means that the image is true.

[0072] Step S204, iteratively optimize the generator and the discriminator through adversarial learning, and obtain a model for improving the clarity of photos of equipment in the computer room after training is completed.

[0073] Specifically, the high-definition pictures and original inspection pictures are input into the discriminator, and the discriminator gives high scores to the original inspection pictures and low scores to the high-definition pictures. The generator aims to obtain high scores from the discriminator and continuously generates high-definition pictures through self-learning, so that the generator and discriminator are iteratively optimized through adversarial learning, gradually improving the quality of the generated high-definition pictures. After the training is completed, the final model for improving the clarity of computer room equipment photos is obtained.

[0074] Further, step S202 includes:

[0075] The low-resolution image is upsampled through the convolutional neural network of the generator to generate a high-resolution image.

[0076] Further, see Figure 2 , after step S103, further comprising the following steps:

[0077] Step S301, inputting the first test data set into the computer room equipment photo clarity improvement model to obtain a high-definition picture; wherein the high-definition picture is output according to a predetermined size;

[0078] Step S302, calculating a quantitative indicator PSNR (Peak Signal-to-Noise Ratio) according to the high-definition picture and the original high-definition picture; wherein PSNR is an indicator for quantifying picture quality, and is used to compare the quality loss between the original high-definition image and the processed high-definition image.

[0079] Step S303, determining whether the quantitative indicator PSNR is within a preset indicator range; if yes, executing step S304; if no, executing step S305;

[0080] Step S304, determining whether the model for improving the clarity of equipment photos in the computer room meets the requirements;

[0081] Step S305: determine that the model for improving the clarity of equipment photos in the computer room does not meet the requirements.

[0082] Further, step S302 includes the following steps:

[0083] Step S401, calculating a mean square error based on the high-definition picture and the original high-definition picture;

[0084] Specifically, the mean square error is calculated according to formula (1):

[0085] MSE=∑((l1(i,j)-l2(i,j)^2) / (M*N) (1)

[0086] Where MSE is the mean square error, l1 is the original HD image, l2 is the HD image, (i, j) is the pixel coordinate, M is the width of the original HD image, and N is the height of the original HD image. The lower the value, the more similar the original HD image and the HD image are, that is, the higher the quality, the width of the original HD image is the same as the width of the HD image, and the height of the original HD image is the same as the height of the HD image.

[0087] Step S402, calculating the quantitative indicator PSNR according to the mean square error and the maximum value of the image pixel value.

[0088] Specifically, the quantitative indicator PSNR is calculated according to formula (2):

[0089] PSNR=10*log10((peak^2) / MSE) (2)

[0090] Here, peak is the maximum value of the image pixel value (usually 255 for 8-bit images).

[0091] Furthermore, after step S305, the method further includes the following steps:

[0092] Step S501, adding multiple inspection pictures to the first sample data set to obtain a second sample data set;

[0093] Step S502, dividing the second sample data set into a second training data set and a second test data set according to a preset ratio;

[0094] Step S503, using the principle of the generative adversarial network model, re-training the generator and the discriminator through the second training data set to obtain a re-trained model for improving the clarity of the equipment room photos;

[0095] Step S504, input the second test data set into the retrained computer room equipment photo clarity improvement model, and calculate the quantitative indicator PSNR until the quantitative indicator PSNR is within the preset indicator range, and the retrained computer room equipment photo clarity improvement model meets the requirements.

[0096] Furthermore, after step S304, the following steps are also included:

[0097] Step S601, obtaining the inspection picture to be detected;

[0098] Step S602, input the inspection picture to be detected into the equipment room equipment photo clarity improvement model to obtain a high-definition inspection picture to be detected;

[0099] Step S603, obtaining high-definition historical inspection pictures;

[0100] Step S604, determining the state of the monitoring device by comparing the high-definition inspection picture to be detected with the high-definition historical inspection picture, and storing the high-definition inspection picture to be detected as a high-definition historical inspection picture.

[0101] Specifically, the images to be inspected are obtained, which are used to monitor and evaluate the status of the equipment and identify potential problems in the equipment status. The images to be inspected are input into the equipment room photo clarity improvement model to improve the image clarity and ensure that the image clarity is high enough to accurately analyze the equipment status and effectively identify potential problems.

[0102] By comparing the high-definition inspection pictures to be tested with the high-definition historical inspection pictures, the changes in the equipment status can be determined. Based on the comparison results, the health status of the monitoring equipment is determined and necessary measures are taken. At the same time, the high-definition inspection pictures to be tested will also be stored as new high-definition historical inspection pictures, so as to provide a more reliable data basis for the accurate recording and subsequent analysis of the equipment status in the computer room during the subsequent analysis, reporting and troubleshooting process.

[0103] Through the above process, the status of the monitoring equipment can be efficiently detected, potential problems can be identified in advance, and necessary maintenance measures can be taken, thereby ensuring the normal operation of the monitoring equipment and extending the service life of the monitoring equipment.

[0104] Embodiment 2:

[0105] Figure 3 Schematic diagram of the AI-based inspection image clarity training device provided in Example 2 of the present invention.

[0106] The first acquisition module is used to collect inspection pictures of equipment in the operation and maintenance room and use them as the first sample data set;

[0107] A first division module, used to divide the first sample data set into a first training data set and a first test data set according to a preset ratio;

[0108] The first adversarial training module is used to utilize the principle of the generative adversarial network model to perform adversarial training on the generator and the discriminator through the first training data set to obtain a model for improving the clarity of photos of equipment in the computer room.

[0109] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the AI-based inspection image clarity training method provided in the above embodiment are implemented.

[0110] An embodiment of the present invention also provides a computer-readable medium having a non-volatile program code executable by a processor, on which a computer program is stored. When the computer program is executed by the processor, the steps of the AI-based inspection image clarity training method of the above embodiment are executed.

[0111] The computer program product provided in the embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can be found in the method embodiment and will not be repeated here.

[0112] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0113] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0114] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0115] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0116] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A training method for inspection image clarity based on AI, characterized in that: The method comprises: Collect inspection pictures of equipment in the operation and maintenance room and use them as the first sample data set; Dividing the first sample data set into a first training data set and a first test data set according to a preset ratio; By utilizing the principle of the generative adversarial network model, the generator and the discriminator are adversarially trained using the first training data set to obtain a model for improving the clarity of photos of equipment in a computer room.

2. The AI-based inspection picture clarity training method according to claim 1 is characterized in that: By using the principle of the generative adversarial network model, the generator and the discriminator are trained adversarially using the first training data set to obtain a model for improving the clarity of photos of equipment in the computer room, including: Performing image resolution degradation processing on the original inspection images in the first training data set to obtain low-resolution images corresponding to the original inspection images; Inputting the low-definition picture into the generator to generate a high-definition picture; Inputting the high-definition picture and the original inspection picture into the discriminator, and outputting an evaluation result, wherein the evaluation result is used to evaluate the authenticity of the generated high-definition picture; The generator and the discriminator are iteratively optimized through adversarial learning, and after training, the clarity improvement model of the equipment room photos is obtained.

3. The AI-based inspection picture clarity training method according to claim 1 is characterized in that: Before dividing the first sample data set into a first training data set and a first test data set according to a preset ratio, the method further includes: The first sample data set is preprocessed, and the preprocessing includes resizing, normalization, and cropping to ensure that the size of the original inspection images in the first training data set remains consistent.

4. The AI-based inspection picture clarity training method according to claim 1 is characterized in that: After performing adversarial training on the generator and the discriminator through the first training data set to obtain a model for improving the clarity of photos of equipment in a computer room, the method further includes: Inputting the first test data set into the computer room equipment photo clarity improvement model to obtain a high-definition picture; Calculate the quantitative indicator PSNR based on the high-definition picture and the original high-definition picture; Determine whether the quantitative indicator PSNR is within a preset indicator range; If yes, it is determined that the model for improving the clarity of the equipment room photos meets the requirements; If not, it is determined that the computer room equipment photo clarity improvement model does not meet the requirements.

5. The AI-based inspection image clarity training method according to claim 4 is characterized in that: The quantitative indicator PSNR is calculated based on the high-definition picture and the original high-definition picture, including: Calculating a mean square error based on the high-definition image and the original high-definition image; The quantitative indicator PSNR is calculated according to the maximum value of the mean square error and the image pixel value.

6. The AI-based inspection image clarity training method according to claim 4 is characterized in that: After determining that the equipment room equipment photo clarity improvement model does not meet the requirements, it also includes: Adding multiple inspection pictures to the first sample data set to obtain a second sample data set; Dividing the second sample data set into a second training data set and a second test data set according to a preset ratio; Using the principle of the generative adversarial network model, the generator and the discriminator are re-trained adversarially using the second training data set to obtain a retrained model for improving the clarity of photos of equipment in the computer room; After inputting the second test data set into the retrained computer room equipment photo clarity improvement model, and calculating the quantitative indicator PSNR, until the quantitative indicator PSNR is within the preset indicator range, the retrained computer room equipment photo clarity improvement model meets the requirements.

7. The AI-based inspection picture clarity training method according to any one of claims 4 to 6, characterized in that: If yes, after determining that the equipment room equipment photo definition improvement model meets the requirements, the method further includes: Get the inspection pictures to be tested; Input the inspection picture to be detected into the equipment room equipment photo clarity improvement model to obtain a high-definition inspection picture to be detected; Get high-definition historical inspection pictures; The state of the monitoring device is determined by comparing the high-definition inspection picture to be detected with the high-definition historical inspection picture, and the high-definition inspection picture to be detected is stored as the high-definition historical inspection picture.

8. An AI-based inspection image clarity training device, characterized in that: The device comprises: The first acquisition module is used to collect inspection pictures of equipment in the operation and maintenance room and use them as the first sample data set; A first division module, used to divide the first sample data set into a first training data set and a first test data set according to a preset ratio; The first adversarial training module is used to utilize the principle of the generative adversarial network model to perform adversarial training on the generator and the discriminator through the first training data set to obtain a model for improving the clarity of photos of equipment in the computer room.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the computer program, the AI-based inspection image clarity training method described in any one of claims 1 to 7 is implemented.

10. A computer readable medium having a non-volatile program code executable by a processor, characterized in that: The program code enables the processor to execute the AI-based inspection image clarity training method described in any one of claims 1 to 7.

Images