Model fine tuning method and system for gray fabric defect over-detection, medium and computer
By collecting and synthesizing training sets of grey fabric products and defect photos, the deep learning model is fine-tuned, which solves the detection inaccurate problem caused by the texture differences of new fabrics, and improves the accuracy and efficiency of automated detection in the textile industry.
Patent Information
- Application Number
- CN202411254358.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-09
- Publication Date
- 2024-12-31
AI Technical Summary
When the texture of the new cloth seed is similar to that of the existing cloth seed, the model's detection ability of the new sample fabric is reduced, and the existing model fine-tuning method ignores the detection ability of the original data or the defective pictures of the new data. Causes inaccurate detection.
By collecting grey cloth products photos and synthetic defect photos, merge them into the model training set, and fine-tuning the adjustment model with this training set, freezing the feature extraction layer and fully connected layer, and conducting 30 to 50 trainings to ensure that the model has new cloths and the ability to identify existing cloths.
The model's learning ability of different types of cloth and defect characteristics is improved, the false detection rate is reduced, manual intervention is reduced, and the accuracy and efficiency of detection is improved. It is suitable for frequent replacement of detection targets.
Smart Images
Figure CN119228746A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of automatic detection of grey cloth defects, and more specifically, to a model fine-tuning method, system, medium and computer for grey cloth defect detection. Background Art
[0002] In recent years, with the development of deep learning technology, the application of computer vision technology based on deep neural networks in the field of industrial defect detection has become more and more extensive. Fabric defect detection is an important task in the textile industry. In the existing technology, a weaving and inspection robot is used, which is an AI robot for automatic defect detection of grey fabric. It can automatically identify defects on grey fabric by analyzing the image of grey fabric while weaving the grey fabric.
[0003] However, factories frequently change fabrics, and sometimes the texture of new fabrics is similar to the defective texture of existing fabrics. This will cause the model trained with existing data to make a large number of false detections for the good background of new fabrics, thus affecting the performance of the automated inspection AI robot. How to effectively solve the problem of new fabrics failing to pass inspection due to texture differences has become an important part of the implementation process. The effective method now is to fine-tune the model to ensure that the detection and recognition capabilities of the new fabrics are taken into account. However, there are some problems with the existing fine-tuning methods. Some methods ignore the detection capabilities of the original data, while some methods ignore the recognition capabilities of new data.
[0004] There are two problems with the existing model fine-tuning strategy: first, it only considers adding new data, which will reduce the model's ability to recognize previous data; second, it only uses good product images in the new data and ignores defective images, resulting in insufficient defect detection capabilities of the model for new data. Summary of the invention
[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a model fine-tuning method, system, medium and computer for grey cloth defect inspection, so as to overcome the shortcomings of the prior art in which the deep learning model is used to detect grey cloth defects, which will lead to insufficient learning samples of the texture of new cloth types and cause incorrect detection.
[0006] The above technical objectives of the present invention are achieved through the following technical solutions: A model fine-tuning method for grey cloth defect inspection, comprising:
[0007] Based on the detection result of the detection mold to be adjusted, judging whether the detection result is over-detected; if so, collecting a first number of photos of good grey cloth products;
[0008] Based on the grey cloth good product photos, synthesizing a second number of grey cloth defect photos;
[0009] The good grey cloth photo, the defective grey cloth photo and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model are combined into a model training set;
[0010] The model to be adjusted is adjusted using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
[0011] Optionally, the model fine-tuning method also includes: before judging whether the detection result is over-inspected based on the detection result of the detection model to be adjusted, first judging whether the cloth type data of the grey cloth is stored in the database; if not, judging whether the detection result is over-inspected based on the detection result of the grey cloth.
[0012] Optionally, collecting a first number of grey cloth good product photos specifically includes: collecting 900-1200 grey cloth good product photos.
[0013] Optionally, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos specifically includes: enhancing the grey cloth good product photos to obtain enhanced photos; extracting contours of the enhanced photos to obtain grey cloth texture, and designing artificial defects based on the grey cloth texture; and synthesizing the grey cloth good product photos and artificial defects to obtain grey cloth defect photos.
[0014] Optionally, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos further includes: performing data augmentation on the grey cloth defect photos to obtain a second number of grey cloth defect photos.
[0015] Optionally, the model to be adjusted is adjusted using the model training set to obtain a grey cloth detection model that can accurately judge grey cloth defects, specifically comprising: freezing the last feature extraction layer and the last fully connected layer of the model to be adjusted; and training the frozen model to be adjusted 30 to 50 times using the model training set to obtain a grey cloth detection model that can accurately judge grey cloth defects.
[0016] Optionally, the frozen model to be adjusted is trained 30 to 50 times using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects, specifically comprising: taking the weight parameters of the model to be adjusted as initial weight parameters, training the model to be adjusted 30 to 50 times using the model training set, adjusting the initial weight parameters in the model to be adjusted, and obtaining a grey cloth detection model capable of accurately judging grey cloth defects, wherein the model to be adjusted is a stable version model.
[0017] A model fine-tuning system for grey cloth defect inspection, comprising:
[0018] An over-inspection judgment module, the over-inspection judgment module is used to judge whether the detection result has an over-inspection phenomenon based on the detection result of the detection module to be adjusted; if so, collect a first number of grey cloth good product photos;
[0019] A photo synthesis module, the photo synthesis module is used to synthesize a second number of grey cloth defect photos based on the grey cloth good product photos;
[0020] A training set merging module, the training set merging module is used to merge the grey cloth good product photos, the grey cloth defect photos and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model into a model training set;
[0021] The model adjustment module is used to adjust the model to be adjusted using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
[0022] A computer-readable storage medium stores a computer program, which implements the steps of the above method when executed by a processor.
[0023] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.
[0024] In summary, the present invention has the following beneficial effects: the merged data set contains stable version data, new fabric good product pictures and generated defect pictures, ensuring that the model can learn the characteristics of different fabric types and defects and improve the generalization ability. By using the stable version data as part of the training set, it is ensured that the model still maintains high recognition ability and low false detection rate for existing fabric types when processing new fabric types. The purpose of the present invention is to solve the problem of inaccurate model recognition due to the texture differences of new fabric types caused by frequent changes in fabric types in weaving factories, and to reduce the need for manual intervention. By automatically triggering the finetune process of the new fabric type model, it is possible to effectively deal with the over-inspection caused by the texture differences of different fabrics. This technology is not only suitable for weaving factories, but can also be applied to other industrial inspection scenarios where the inspection targets are frequently changed, thereby improving the accuracy and efficiency of the inspection and reducing labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 It is a flow chart of a model fine-tuning method for grey cloth defect inspection of the present invention;
[0026] Figure 2 This is a structural diagram of a model fine-tuning system for grey cloth defect inspection according to the present invention;
[0027] Figure 3 1 is a diagram showing the internal structure of a computer device in an embodiment of the present invention.
[0028] In the figure: 1. Inspection judgment module; 2. Photo synthesis module; 3. Training set merging module; 4. Model adjustment module. DETAILED DESCRIPTION
[0029] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. Several embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein.
[0030] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like 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 it can be indirectly connected through an intermediate medium, or it can be the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. The terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.
[0031] In the present invention, unless otherwise clearly specified and limited, the first feature being "above" or "below" the second feature may include the first and second features being in direct contact, or may include the first and second features being in contact not directly but through another feature between them. Moreover, the first feature being "above", "above" and "above" the second feature includes the first feature being directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. The first feature being "below", "below" and "below" the second feature includes the first feature being directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature. The terms "vertical", "horizontal", "left", "right", "above", "below" and similar expressions are for illustrative purposes only, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.
[0032] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments.
[0033] The present invention provides a model fine-tuning method for grey cloth defect inspection, such as Figure 1 As shown, including:
[0034] S1. Based on the detection result of the detection mold to be adjusted, determine whether the detection result is over-detected; if so, collect a first number of photos of good grey cloth products;
[0035] S2. synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos;
[0036] S3, combining the grey cloth good product photos, the grey cloth defect photos and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model into a model training set;
[0037] S4. Using the model training set to adjust the model to be adjusted, to obtain a grey cloth detection model that can accurately determine grey cloth defects.
[0038] In the traditional field of grey cloth online monitoring technology, due to the complexity and diversity of the texture of grey cloth, the deep learning model is used to automatically detect the defects of grey cloth, which is easily affected by the texture of the grey cloth, and the texture of the grey cloth is mistaken for the defect of the grey cloth, resulting in a large number of erroneous defects being identified when detecting a certain type of cloth. This situation is called over-inspection. In order to enable the model to accurately identify the texture of the new grey cloth and not affect the grey cloth's ability to detect the defects of the grey cloth that has been trained in the past, the present application proposes a model fine-tuning method for over-inspection of grey cloth defects. First, when the grey cloth textile equipment changes the type of cloth, first determine whether the cloth type is stored in the database. If it is not stored in the database, it is necessary to first determine the detection results of the grey cloth to see if there is an over-inspection situation. If so, the subsequent steps of this application are automatically executed. The subsequent steps specifically include: collecting good photos of grey fabric, which refers to good photos of grey fabric being woven by textile equipment, selecting good photos without defects through manual screening, and then synthesizing the defects of grey fabric on the good photos to form grey fabric defect photos. Grey fabric defect photos can enable the model to identify the defects of grey fabric on the basis of the grey fabric texture. In order to avoid affecting the accuracy of the model's detection of the original grey fabric texture during the model training process, it is also necessary to select a part of the stable version data of the model to be adjusted to form a training set together, where the stable version data refers to the data set that has been used in the production environment and has been fully verified. These data sets contain confirmed good and defective pictures, and the performance of the model on these data is stable and reliable. By forming a training set with the stable version data of the model, good grey fabric photos, and grey fabric defect photos and inputting them into the model for training, some weight parameters in the model can be adjusted so that the model can improve the detection ability of new fabric types while maintaining the existing detection ability, and avoid over-inspection when detecting new fabric types.
[0039] In summary, the purpose of the present invention is to solve the problem of inaccurate model recognition due to the texture differences of new fabrics caused by frequent changes in fabric types in weaving factories, and to reduce the need for manual intervention. By automatically triggering the fine-tuning process of the new fabric model, we can effectively deal with the over-inspection situation caused by the texture differences of different fabrics. This technology is not only suitable for weaving factories, but can also be applied to other industrial inspection scenarios where the inspection target is frequently changed, thereby improving the accuracy and efficiency of the inspection and reducing labor costs.
[0040] In one embodiment, the model fine-tuning method also includes: before judging whether the detection result is over-inspected based on the detection result of the detection model to be adjusted, first judging whether the cloth type data of the grey cloth is stored in the database; if not, judging whether the detection result is over-inspected based on the detection result of the grey cloth.
[0041] In actual use, the above method can quickly detect whether it is a new fabric type without manual intervention, thereby improving production efficiency. It can also immediately trigger subsequent detection and adjustment processes when new fabric types are put online, reducing production downtime caused by fabric type changes.
[0042] In one embodiment, collecting a first number of grey cloth good product photos specifically includes: collecting 900-1200 grey cloth good product photos.
[0043] In actual use, according to experimental results, when there are hundreds of thousands of model training data, the applicant found that collecting 1,000 good product images can completely suppress the false detection caused by texture differences when new fabrics are put online. If the number of good product images collected is too small, false detection cannot be effectively reduced. If too many good product images are collected, the training time will increase.
[0044] In one embodiment, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos specifically includes: enhancing the grey cloth good product photos to obtain enhanced photos; extracting contours of the enhanced photos to obtain grey cloth texture, and designing artificial defects based on the grey cloth texture; and synthesizing the grey cloth good product photos and artificial defects to obtain grey cloth defect photos.
[0045] In actual use, the applicant found that adding only good product pictures would inhibit the model's ability to identify defects in new types of fabrics. Therefore, it is necessary to add defective photos of new types of grey fabrics to train together while increasing the number of good product photos, so that the model can maintain accurate defect detection capabilities without failing inspections.
[0046] In one embodiment, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos further includes: performing data augmentation on the grey cloth defect photos to obtain a second number of grey cloth defect photos.
[0047] In order to increase the number of grey fabric defect photos, it is necessary to perform data augmentation on the grey fabric photos. By simulating real environment interference and defect synthesis, the model's generalization ability for defects in different environments and fabric types can be improved, and the robustness of the model can be enhanced. The synthesized defect images can help the model better identify and recall real defects on the fabric and reduce the missed detection rate.
[0048] In one embodiment, the method of adjusting the model to be adjusted by using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects specifically includes: freezing the last feature extraction layer and the last fully connected layer of the model to be adjusted; and training the frozen model to be adjusted 30 to 50 times by using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects.
[0049] In actual use, freezing the last feature extraction layer and the last fully connected layer of the model to be adjusted can ensure that the new model maintains a high recognition ability for existing fabrics while processing new fabrics. By freezing the last feature extraction layer and the fully connected layer, the stability of the existing model is maintained and the training process is accelerated. By using the weights of the stable version of the model as the initialization weights, the training efficiency is improved and the training time is shortened.
[0050] In one embodiment, the frozen model to be adjusted is trained 30 to 50 times using the model training set to obtain a grey cloth detection model that can accurately judge grey cloth defects, specifically including: using the weight parameters of the model to be adjusted as initial weight parameters, training the model to be adjusted 30 to 50 times using the model training set, adjusting the initial weight parameters in the model to be adjusted, and obtaining a grey cloth detection model that can accurately judge grey cloth defects, wherein the model to be adjusted is a stable version model. In this application, a stable version model refers to a model that can be fully verified during the production process, and the detection results are accurate and reliable.
[0051] In summary, by adding 10% of the stable version data to the model training set, the present application can maintain the model's stable recognition of the original data. By adding photos of good grey cloth and photos of grey cloth defects, the model can maintain the model's ability to detect grey cloth defects while recognizing the texture of the grey cloth. And only using 10% of the stable version data can effectively improve the adjustment speed of the model and avoid reducing the adjustment efficiency of the model due to too much stable version data. In practical applications, the method of the present application can be automatically executed to determine whether the model needs fine-tuning based on the detection results of the model. If so, the method is automatically executed to fine-tune the model, reducing the need for manual intervention and improving the accuracy and efficiency of detection.
[0052] like Figure 2As shown, the present invention also provides a model fine-tuning system for grey cloth defect inspection, comprising:
[0053] An over-inspection judgment module 1 is used to judge whether the detection result has an over-inspection phenomenon based on the detection result of the detection module to be adjusted; if so, collect a first number of grey cloth good product photos;
[0054] A photo synthesis module 2, the photo synthesis module 2 is used to synthesize a second number of grey cloth defect photos based on the grey cloth good product photos;
[0055] A training set merging module 3, wherein the training set merging module 3 is used to merge the grey cloth good product photos, the grey cloth defect photos and the pre-selected stable recognition data corresponding to the detection model to be adjusted into a model training set;
[0056] The model adjustment module 4 is used to adjust the model to be adjusted using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
[0057] For the specific definition of a model fine-tuning system for grey cloth defect inspection, please refer to the definition of a model fine-tuning method for grey cloth defect inspection mentioned above, which will not be repeated here. Each module in the above-mentioned model fine-tuning system for grey cloth defect inspection can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0058] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the computer program is executed by the processor, a model fine-tuning method for defect inspection of grey cloth is implemented.
[0059] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0060] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:
[0061] S1. Based on the detection result of the detection mold to be adjusted, determine whether the detection result is over-detected; if so, collect a first number of photos of good grey cloth products;
[0062] S2. synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos;
[0063] S3, combining the grey cloth good product photos, the grey cloth defect photos and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model into a model training set;
[0064] S4. Using the model training set to adjust the model to be adjusted, to obtain a grey cloth detection model that can accurately determine grey cloth defects.
[0065] In one embodiment, the model fine-tuning method also includes: before judging whether the detection result is over-inspected based on the detection result of the detection model to be adjusted, first judging whether the cloth type data of the grey cloth is stored in the database; if not, judging whether the detection result is over-inspected based on the detection result of the grey cloth.
[0066] In one embodiment, collecting a first number of grey cloth good product photos specifically includes: collecting 900-1200 grey cloth good product photos.
[0067] In one embodiment, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos specifically includes: enhancing the grey cloth good product photos to obtain enhanced photos; extracting contours of the enhanced photos to obtain grey cloth texture, and designing artificial defects based on the grey cloth texture; and synthesizing the grey cloth good product photos and artificial defects to obtain grey cloth defect photos.
[0068] In one embodiment, synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos further includes: performing data augmentation on the grey cloth defect photos to obtain a second number of grey cloth defect photos.
[0069] In one embodiment, the method of adjusting the model to be adjusted by using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects specifically includes: freezing the last feature extraction layer and the last fully connected layer of the model to be adjusted; and training the frozen model to be adjusted 30 to 50 times by using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects.
[0070] In one embodiment, the frozen model to be adjusted is trained 30 to 50 times using the model training set to obtain a grey cloth detection model capable of accurately judging grey cloth defects, specifically comprising: taking the weight parameters of the model to be adjusted as initial weight parameters, training the model to be adjusted 30 to 50 times using the model training set, adjusting the initial weight parameters in the model to be adjusted, and obtaining a grey cloth detection model capable of accurately judging grey cloth defects, wherein the model to be adjusted is a stable version model.
[0071] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0072] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0073] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.
Claims
1. A model fine-tuning method for grey cloth defect inspection, characterized in that: include: Based on the detection result of the detection module to be adjusted, determining whether the detection result has an over-detection phenomenon; If yes, then collect a first number of good grey cloth photos; Based on the grey cloth good product photos, synthesizing a second number of grey cloth defect photos; The good grey cloth photo, the defective grey cloth photo and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model are combined into a model training set; The model to be adjusted is adjusted using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
2. A model fine-tuning method for grey cloth defect inspection according to claim 1, characterized in that: The model fine-tuning method further includes: Before judging whether the detection result is over-tested based on the detection result of the detection mold to be adjusted, first judge whether the cloth type data of the grey cloth is stored in the database. If not, judge whether the detection result is over-tested based on the detection result of the grey cloth.
3. The model fine-tuning method for grey cloth defect inspection according to claim 1, characterized in that: The collecting of the first number of grey cloth good product photos specifically includes: Collect 900-1200 photos of good grey fabric.
4. The model fine-tuning method for grey cloth defect inspection according to claim 1, characterized in that: The synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos specifically includes: Performing enhancement processing on the grey cloth good product photo to obtain an enhanced photo; Extracting contours from the enhanced photo to obtain a grey fabric texture, and designing artificial defects based on the grey fabric texture; The grey cloth good product photo and the artificial defect photo are synthesized to obtain the grey cloth defect photo.
5. The model fine-tuning method for grey cloth defect inspection according to claim 4 is characterized in that: The step of synthesizing a second number of grey cloth defect photos based on the grey cloth good product photos further includes: Data augmentation is performed on the grey cloth defect photos to obtain a second number of grey cloth defect photos.
6. The model fine-tuning method for grey cloth defect inspection according to claim 1, characterized in that: The method of using the model training set to adjust the model to be adjusted to obtain a grey cloth detection model capable of accurately judging grey cloth defects specifically includes: Freeze the last feature extraction layer and the last fully connected layer of the model to be adjusted; The frozen model to be adjusted is trained 30 to 50 times using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
7. A model fine-tuning method for grey cloth defect inspection according to claim 6, characterized in that: The method of using the model training set to train the frozen model to be adjusted 30 to 50 times to obtain a grey cloth detection model capable of accurately judging grey cloth defects specifically includes: The weight parameters of the model to be adjusted are used as the initial weight parameters, and the model to be adjusted is trained 30 to 50 times using the model training set. The initial weight parameters in the model to be adjusted are adjusted to obtain a grey cloth detection model that can accurately judge grey cloth defects, wherein the model to be adjusted is a stable version model.
8. A model fine-tuning system for grey cloth defect inspection, characterized in that: include: An over-test judgment module, which is used to judge whether there is an over-test phenomenon in the test result based on the test result of the test module to be adjusted; If yes, then collect a first number of good grey cloth photos; A photo synthesis module, the photo synthesis module is used to synthesize a second number of grey cloth defect photos based on the grey cloth good product photos; A training set merging module, the training set merging module is used to merge the grey cloth good product photos, the grey cloth defect photos and the pre-selected stable recognition data corresponding to the to-be-adjusted detection model into a model training set; The model adjustment module is used to adjust the model to be adjusted using the model training set to obtain a grey cloth detection model that can accurately determine grey cloth defects.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.