Sample chart normalization detection method, system and equipment and storage medium
By training the sample clothing standard detection model, the standardization of the front, side and back views of the sample clothing is detected, the problem of lack of unified standards in the existing technology is solved, the quality of sample clothing photography and review efficiency is improved, and more standard pictures are provided .
Patent Information
- Application Number
- CN202410979014.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing sample clothing shooting process lacks unified shooting standards, resulting in messy pictures, increased review difficulty, and extended time, which is not conducive to the maintenance and management of sample clothing information.
By obtaining the training sample set, the neural network is trained, and the sample garment specification detection model is generated, which detects whether the front, side and back of the sample garment meet the specifications, and determines whether the pictures are uploaded.
It improves the standardization and review efficiency of sample clothes pictures, and provides more standardized pictures for downstream review process.
Smart Images

Figure CN118968215A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of clothing management, and more specifically, to a method, system, device and storage medium for detecting the standardization of a sample clothing image. Background Art
[0002] At present, in the process of shooting sample garments, most of the shooting processes are as follows: the photographer uses the shooting tool to shoot according to the requirements, then uploads it to the system, and then conducts manual review of the sample. This process is relatively simple to implement and can complete the sample shooting and subsequent management process to a certain extent.
[0003] However, the process still has the following problems: 1. Due to the lack of unified shooting standards and specifications, the final pictures will be messy, increasing the difficulty of review; 2. The randomness of the pictures increases the difficulty and time of the review process; 3. Non-standard pictures are not conducive to the maintenance and management of related sample information in the later stage. Due to the existence of these three problems, the manual review of sample clothes will have problems such as untimely supervision, inconsistent standards, and reduced timeliness. Summary of the invention
[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a sample clothing image standardization detection method, system, equipment and storage medium, which have the functional advantages of being more efficient, more reasonable and effective in standardizing the quality of sample clothing image shooting, and providing more standardized images for the downstream review process.
[0005] The above technical objectives of the present invention are achieved through the following technical solutions:
[0006] A method for detecting the standardization of a sample garment drawing, comprising:
[0007] Obtain a training sample set, a front image of a sample garment, a side image of a sample garment, and a back image of a sample garment;
[0008] The training sample set is used to train the neural network to obtain a sample clothing standard detection model;
[0009] Inputting the sample clothing front view, sample clothing side view and sample clothing back view into the sample clothing detection standard model to obtain the detection result;
[0010] Determine whether to upload the front view, side view and back view of the sample garment according to the detection result.
[0011] Optionally, the training sample set includes: overall clothing sample images, partial clothing sample images and non-clothing sample images; the overall clothing sample images are positive samples; the partial clothing sample images and non-clothing sample images are negative samples.
[0012] Optionally, the step of using the training sample set to train a neural network to obtain a sample clothing specification detection model includes:
[0013] Pre-training the neural network using the positive samples and negative samples;
[0014] Dividing the positive samples into a first test set and a first scene set, and dividing the negative samples into a second test set and a second scene set;
[0015] Fine-tune the neural network using the first test set, the second test set, and the trajectory generated by the pre-trained neural network until the training result reaches a first preset threshold;
[0016] The neural network is fine-tuned using the first scene set, the second scene set, and the trajectory generated by the fine-tuned neural network until the training result reaches a second preset threshold, thereby obtaining a sample clothing specification detection model.
[0017] Optionally, the pre-training the neural network using the positive samples and the negative samples includes:
[0018] Adding corresponding sample labels to the positive samples and negative samples;
[0019] The positive samples and negative samples are used to train the neural network to obtain corresponding training detection categories, a loss function is calculated according to the training detection categories and corresponding sample labels, and the neural network parameters are updated.
[0020] Optionally, the neural network is a YOLO neural network; the sample clothing standard detection model includes: an input module, a Backbone module, a Neck module and a Head module; the front view, side view and back view of the sample clothing are input into the sample clothing detection standard model to obtain a detection result, including:
[0021] Inputting the sample garment front view, the sample garment side view and the sample garment back view to an input module, wherein the input module is used to scale and crop the sample garment front view, the sample garment side view and the sample garment back view;
[0022] The Backbone module is used to extract features from the scaled and cropped front image, side image and back image of the sample garment to obtain a sample garment front feature vector, a sample garment side feature vector and a sample garment back feature vector;
[0023] The Neck module is used to perform feature fusion on the feature vector of the front side of the sample, the feature vector of the side of the sample and the feature vector of the back of the sample to obtain a feature map of the front side of the sample, a feature map of the side of the sample and a feature map of the back of the sample;
[0024] The Head module is used to predict and classify the front feature image, side feature image and back feature image of the sample clothing, and judge whether the front image, side image and back image of the sample clothing are standardized according to the confidence level.
[0025] Optionally, judging whether the sample clothing front view, the sample clothing side view and the sample clothing back view meet the specification according to the confidence level includes:
[0026] If it is determined according to the confidence level that the sample clothing front image is the front of the sample clothing and the human body / mannequin in the sample clothing front image has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing front image meets the specification;
[0027] If it is determined according to the confidence level that the sample clothing side view is a sample clothing side view and the human body / mannequin in the sample clothing side view has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing side view meets the specification;
[0028] When it is determined based on the confidence level that the sample back image is the back of the sample and the human body / mannequin in the sample back image has an upper body, a lower body, arms, shoulders and a whole body, it is determined that the sample back image meets the specifications.
[0029] Optionally, also include:
[0030] After the detection result of the sample front image, the sample side image or the sample back image is that the number of times it does not meet the specifications exceeds a third preset threshold, the sample front image, the sample side image or the sample back image is uploaded.
[0031] A sample clothing standardization detection system, characterized by comprising:
[0032] An acquisition module is used to acquire a training sample set, a front view of a sample garment, a side view of a sample garment, and a back view of a sample garment;
[0033] A training module, used to train a neural network using the training sample set to obtain a sample clothing standard detection model;
[0034] A detection module, used for inputting the sample clothing front view, the sample clothing side view and the sample clothing back view into the sample clothing detection standard model to obtain a detection result;
[0035] The judgment module is used to judge whether to upload the front view, side view and back view of the sample clothing according to the detection result.
[0036] 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.
[0037] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.
[0038] In summary, the present invention has the following beneficial effects: a neural network is trained using a training sample set to obtain a sample clothing standard detection model, so as to detect the photographed front view, side view and back view of the sample clothing, determine whether the front view, side view and back view of the sample clothing meet the standards, and upload the front view, side view and / or back view of the sample clothing that meet the standards to the system. Compared with the manual review of whether the front view, side view and back view of the sample clothing meet the standards in the prior art, this is more efficient, more reasonable and effective in standardizing the quality of sample clothing picture shooting, and provides more standardized pictures for the downstream review process. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 It is a schematic flow chart of the method provided by the present invention;
[0040] Figure 2 It is a structural block diagram of the sample clothing standardization detection system provided by the present invention;
[0041] Figure 3 It is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION
[0042] 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.
[0043] 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.
[0044] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments.
[0045] The present invention provides a method for detecting the standardization of a sample clothing image. Figure 1 As shown, including:
[0046] Step 1, obtain a training sample set, a front view of a sample garment, a side view of a sample garment, and a back view of a sample garment;
[0047] Step 2: Using the training sample set to train the neural network to obtain a sample clothing standard detection model;
[0048] Step 3: input the sample clothing front view, sample clothing side view and sample clothing back view into the sample clothing detection standard model to obtain the detection result;
[0049] Step 4: Determine whether to upload the front view, side view and back view of the sample garment according to the detection result.
[0050] In practical applications, the training sample set can be obtained from the Internet by photographing the sample clothes or using a crawler. The front, side and back images of the sample clothes are obtained by photographing. The neural network is trained using the training sample set to obtain a sample clothes standard detection model, so as to detect the photographed front, side and back images of the sample clothes, and determine whether the front, side and back images of the sample clothes meet the specifications. The front, side and / or back images of the sample clothes that meet the specifications are uploaded to the system. Compared with the manual review of whether the front, side and back images of the sample clothes meet the specifications in the prior art, this method is more efficient, more reasonable and effective in standardizing the quality of sample clothes picture shooting, and provides more standard images for the downstream review process.
[0051] Furthermore, the training sample set includes: overall clothing sample images, partial clothing sample images and non-clothing sample images; the overall clothing sample images are positive samples; the partial clothing sample images and non-clothing sample images are negative samples.
[0052] In practical applications, the overall image of sample clothes includes: front view, side view and background view of multiple sample clothes; non-sample clothes images include: landscape images and animal images, etc. The partial image of sample clothes and non-sample clothes images are added to the training sample set as negative samples, which can achieve better training effect, enhance the generalization and robustness of the sample clothes specification detection model, and effectively reduce the misjudgment of incorrect sample clothes images by the sample clothes specification detection model.
[0053] Furthermore, the method of using the training sample set to train a neural network to obtain a sample clothing specification detection model includes:
[0054] Pre-training the neural network using the positive samples and negative samples;
[0055] Dividing the positive samples into a first test set and a first scene set, and dividing the negative samples into a second test set and a second scene set;
[0056] Fine-tune the neural network using the first test set, the second test set, and the trajectory generated by the pre-trained neural network until the training result reaches a first preset threshold;
[0057] The neural network is fine-tuned using the first scene set, the second scene set, and the trajectory generated by the fine-tuned neural network until the training result reaches a second preset threshold, thereby obtaining a sample clothing specification detection model.
[0058] In practical applications, the neural network is first pre-trained with the positive samples and negative samples, and can be divided into a first test set and a first scene set according to the positive samples, that is, the shooting conditions of the overall sample picture (such as the shooting scene and the sample style, etc.), and the negative samples can also be divided into a second test set and a second scene set according to the shooting scene and the shooting object, and then the first test set and the second test set are used to fine-tune the pre-trained neural network to obtain a fine-tuned neural network, and the first scene set and the second scene set are used to fine-tune the fine-tuned neural network to achieve more targeted training of the neural network, thereby improving the detection accuracy of the sample specification detection model. In the present application, the first preset threshold and the second preset threshold are both 95%, that is, the accuracy of the detection result is required to reach 95% to ensure the detection accuracy of the sample specification detection model.
[0059] Further, the using the positive samples and the negative samples to pre-train the neural network includes:
[0060] Adding corresponding sample labels to the positive samples and negative samples;
[0061] The positive samples and negative samples are used to train the neural network to obtain corresponding training detection categories, a loss function is calculated according to the training detection categories and corresponding sample labels, and the neural network parameters are updated.
[0062] In practical applications, corresponding sample labels are added to positive samples, such as adding front sample labels to front clothing samples, adding side sample labels to side clothing samples, and adding corresponding marks to the five areas of the upper body, lower body, arms, shoulders and whole body in the clothing samples. Other sample labels are added to negative samples to facilitate training of the neural network.
[0063] Further, the neural network is a YOLO neural network; the sample clothing standard detection model includes: an input module, a Backbone module, a Neck module and a Head module; the sample clothing front view, the sample clothing side view and the sample clothing back view are input into the sample clothing detection standard model to obtain a detection result, including:
[0064] Inputting the sample garment front view, the sample garment side view and the sample garment back view to an input module, wherein the input module is used to scale and crop the sample garment front view, the sample garment side view and the sample garment back view;
[0065] The Backbone module is used to extract features from the scaled and cropped front image, side image and back image of the sample garment to obtain a sample garment front feature vector, a sample garment side feature vector and a sample garment back feature vector;
[0066] The Neck module is used to perform feature fusion on the feature vector of the front side of the sample, the feature vector of the side of the sample and the feature vector of the back of the sample to obtain a feature map of the front side of the sample, a feature map of the side of the sample and a feature map of the back of the sample;
[0067] The Head module is used to predict and classify the front feature image, side feature image and back feature image of the sample clothing, and judge whether the front image, side image and back image of the sample clothing are standardized according to the confidence level.
[0068] In practical applications, the input module scales and crops the input sample front image, sample side image and sample back image to make the image meet the input requirements of the Backbone module. The Backbone module uses the Focus structure to slice the image and use the convolution kernel to convolve to obtain the feature vector. Secondly, the CSP structure is used in the backbone network to map the features of the base layer into two parts, and then merge them across stages and levels to reduce the repeated calculation of gradient information and ensure accuracy. The Neck module uses the FPN and PAN network structures to process features and uses the CSP2 structure to fuse the feature vectors of the image. The Head module uses CIOU_Loss as the loss function of the Boundingbox, because CIOU_Loss contains influencing factors and involves groundtruth information. When testing reasoning, there is no groundtruth, so the weighted nms method is also used in the Head module to better predict and classify the sample front feature map, sample side feature map and sample back feature map.
[0069] Further, judging whether the front view, side view and back view of the sample garment meet the specification according to the confidence level includes:
[0070] If it is determined according to the confidence level that the sample clothing front image is the front of the sample clothing and the human body / mannequin in the sample clothing front image has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing front image meets the specification;
[0071] If it is determined according to the confidence level that the sample clothing side view is a sample clothing side view and the human body / mannequin in the sample clothing side view has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing side view meets the specification;
[0072] When it is determined based on the confidence level that the sample back image is the back of the sample and the human body / mannequin in the sample back image has an upper body, a lower body, arms, shoulders and a whole body, it is determined that the sample back image meets the specifications.
[0073] In actual applications, when it is determined that the front image of the sample clothing is the front of the sample clothing, but lacks any one or more regions of the upper torso, lower torso, arms, shoulders and whole body, the front image of the sample clothing is determined to be non-compliant with the specifications, and a prompt is given that one / more regions are missing from the sample clothing. When it is determined that the front image of the sample clothing is the side image of the sample clothing or the back image of the sample clothing, the front image of the sample clothing is also determined to be non-compliant with the specifications, and a prompt is given that the front image is not detected. When it is determined that the front image of the sample clothing is not a sample clothing, the front image of the sample clothing is determined to be non-compliant with the specifications, and a prompt is given that the sample clothing image is not detected. When it is determined that the side image of the sample clothing is the side image of the sample clothing, but lacks any one or more regions of the upper torso, lower torso, arms, shoulders and whole body, the side image of the sample clothing is determined to be non-compliant with the specifications, and a prompt is given that one / more regions are missing from the sample clothing. When the side view of the garment is the front or back of the garment sample, the side view of the garment sample is also determined to be non-compliant with the specification, and a prompt is given that the side view is not detected; when the side view of the garment sample is determined to be non-sample garment, the side view of the garment sample is determined to be non-compliant with the specification, and a prompt is given that the sample garment image is not detected; when the back view of the garment sample is determined to be the back of the garment sample, but any one or more regions of the upper torso, lower torso, arms, shoulders and whole body are missing, the back view of the garment sample is determined to be non-compliant with the specification, and a prompt is given that one / more regions are missing from the garment sample; when the back view of the garment sample is determined to be the front or side of the garment sample, the back view of the garment sample is also determined to be non-compliant with the specification, and a prompt is given that the back view is not detected; when the back view of the garment sample is determined to be non-sample garment, the back view of the garment sample is determined to be non-compliant with the specification, and a prompt is given that the sample garment image is not detected.
[0074] Furthermore, it also includes:
[0075] After the detection result of the sample front image, the sample side image or the sample back image is that the number of times it does not meet the specifications exceeds a third preset threshold, the sample front image, the sample side image or the sample back image is uploaded.
[0076] In actual applications, in order to prevent the situation where the front view, side view or back view of a sample that meets the specifications cannot be uploaded after the sample specification detection model makes a misjudgment, the front view, side view or back view of the sample can be successfully uploaded after the number of times the detection result of the front view of the sample does not meet the specifications exceeds the third preset threshold, the number of times the detection result of the side view of the sample does not meet the specifications exceeds the third preset threshold, or the number of times the detection result of the back view of the sample does not meet the specifications exceeds the third preset threshold. In the present application, the third preset threshold is 3. In other embodiments, the third preset threshold can be increased or decreased according to actual needs.
[0077] The method for detecting the standardization of sample images of the present invention uses a training sample set to train a neural network to obtain a sample image standardization detection model, so as to detect the photographed front image, side image and back image of the sample image, determine whether the front image, side image and back image of the sample image meet the standards, and upload the front image, side image and / or back image of the sample image that meet the standards to the system. Compared with the manual review of whether the front image, side image and back image of the sample image meet the standards in the prior art, this method is more efficient, more reasonable and effective in regulating the shooting quality of sample images, and provides more standard images for the downstream review process.
[0078] like Figure 2 As shown, the present invention also provides a sample clothing standardization detection system, comprising:
[0079] An acquisition module 10 is used to acquire a training sample set, a front view of a sample garment, a side view of a sample garment, and a back view of a sample garment;
[0080] A training module 20, for training a neural network using the training sample set to obtain a sample clothing standard detection model;
[0081] A detection module 30, for inputting the sample clothing front view, the sample clothing side view and the sample clothing back view into a sample clothing detection standard model to obtain a detection result;
[0082] The judgment module 40 is used to judge whether to upload the front view, side view and back view of the sample garment according to the detection result.
[0083] For the specific definition of the sample drawing standardization detection system, please refer to the definition of the sample drawing standardization detection method above, which will not be repeated here. The various modules of the above-mentioned sample drawing standardization detection system can be implemented in whole or in part by software, hardware and their combination. 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.
[0084] 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. Among them, 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. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for detecting the standardization of a sample clothing image is implemented.
[0085] 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.
[0086] In one embodiment, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0087] Obtain a training sample set, a front image of a sample garment, a side image of a sample garment, and a back image of a sample garment;
[0088] The training sample set is used to train the neural network to obtain a sample clothing standard detection model;
[0089] Inputting the sample clothing front view, sample clothing side view and sample clothing back view into the sample clothing detection standard model to obtain the detection result;
[0090] Determine whether to upload the front view, side view and back view of the sample garment according to the detection result.
[0091] In one embodiment, the training sample set includes: overall clothing sample images, partial clothing sample images and non-clothing sample images; the overall clothing sample images are positive samples; the partial clothing sample images and non-clothing sample images are negative samples.
[0092] In one embodiment, the step of using the training sample set to train a neural network to obtain a sample clothing specification detection model includes:
[0093] Pre-training the neural network using the positive samples and negative samples;
[0094] Dividing the positive samples into a first test set and a first scene set, and dividing the negative samples into a second test set and a second scene set;
[0095] Fine-tune the neural network using the first test set, the second test set, and the trajectory generated by the pre-trained neural network until the training result reaches a first preset threshold;
[0096] The neural network is fine-tuned using the first scene set, the second scene set, and the trajectory generated by the fine-tuned neural network until the training result reaches a second preset threshold, thereby obtaining a sample clothing specification detection model.
[0097] In one embodiment, the pre-training the neural network using the positive samples and the negative samples includes:
[0098] Adding corresponding sample labels to the positive samples and negative samples;
[0099] The positive samples and negative samples are used to train the neural network to obtain corresponding training detection categories, a loss function is calculated according to the training detection categories and corresponding sample labels, and the neural network parameters are updated.
[0100] In one embodiment, the neural network is a YOLO neural network; the sample clothing specification detection model includes: an input module, a Backbone module, a Neck module and a Head module; the front view, side view and back view of the sample clothing are input into the sample clothing detection specification model to obtain a detection result, including:
[0101] Inputting the sample garment front view, the sample garment side view and the sample garment back view to an input module, wherein the input module is used to scale and crop the sample garment front view, the sample garment side view and the sample garment back view;
[0102] The Backbone module is used to extract features from the scaled and cropped front image, side image and back image of the sample garment to obtain a sample garment front feature vector, a sample garment side feature vector and a sample garment back feature vector;
[0103] The Neck module is used to perform feature fusion on the feature vector of the front side of the sample, the feature vector of the side of the sample and the feature vector of the back of the sample to obtain a feature map of the front side of the sample, a feature map of the side of the sample and a feature map of the back of the sample;
[0104] The Head module is used to predict and classify the front feature image, side feature image and back feature image of the sample clothing, and judge whether the front image, side image and back image of the sample clothing are standardized according to the confidence level.
[0105] In one embodiment, judging whether the front view, side view and back view of the sample garment meet the specification according to the confidence level includes:
[0106] If it is determined according to the confidence level that the sample clothing front image is the front of the sample clothing and the human body / mannequin in the sample clothing front image has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing front image meets the specification;
[0107] If it is determined according to the confidence level that the sample clothing side view is a sample clothing side view and the human body / mannequin in the sample clothing side view has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing side view meets the specification;
[0108] When it is determined based on the confidence level that the sample back image is the back of the sample and the human body / mannequin in the sample back image has an upper body, a lower body, arms, shoulders and a whole body, it is determined that the sample back image meets the specifications.
[0109] In one embodiment, it further includes:
[0110] After the detection result of the sample front image, the sample side image or the sample back image is that the number of times it does not meet the specifications exceeds a third preset threshold, the sample front image, the sample side image or the sample back image is uploaded.
[0111] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0112] Obtain a training sample set, a front image of a sample garment, a side image of a sample garment, and a back image of a sample garment;
[0113] The training sample set is used to train the neural network to obtain a sample clothing standard detection model;
[0114] Inputting the sample clothing front view, sample clothing side view and sample clothing back view into the sample clothing detection standard model to obtain the detection result;
[0115] Determine whether to upload the front view, side view and back view of the sample garment according to the detection result.
[0116] In one embodiment, the training sample set includes: overall clothing sample images, partial clothing sample images and non-clothing sample images; the overall clothing sample images are positive samples; the partial clothing sample images and non-clothing sample images are negative samples.
[0117] In one embodiment, the step of using the training sample set to train a neural network to obtain a sample clothing specification detection model includes:
[0118] Pre-training the neural network using the positive samples and negative samples;
[0119] Dividing the positive samples into a first test set and a first scene set, and dividing the negative samples into a second test set and a second scene set;
[0120] Fine-tune the neural network using the first test set, the second test set, and the trajectory generated by the pre-trained neural network until the training result reaches a first preset threshold;
[0121] The neural network is fine-tuned using the first scene set, the second scene set, and the trajectory generated by the fine-tuned neural network until the training result reaches a second preset threshold, thereby obtaining a sample clothing specification detection model.
[0122] In one embodiment, the pre-training the neural network using the positive samples and the negative samples includes:
[0123] Adding corresponding sample labels to the positive samples and negative samples;
[0124] The positive samples and negative samples are used to train the neural network to obtain corresponding training detection categories, a loss function is calculated according to the training detection categories and corresponding sample labels, and the neural network parameters are updated.
[0125] In one embodiment, the neural network is a YOLO neural network; the sample clothing specification detection model includes: an input module, a Backbone module, a Neck module and a Head module; the front view, side view and back view of the sample clothing are input into the sample clothing detection specification model to obtain a detection result, including:
[0126] Inputting the sample garment front view, the sample garment side view and the sample garment back view to an input module, wherein the input module is used to scale and crop the sample garment front view, the sample garment side view and the sample garment back view;
[0127] The Backbone module is used to extract features from the scaled and cropped front image, side image and back image of the sample garment to obtain a sample garment front feature vector, a sample garment side feature vector and a sample garment back feature vector;
[0128] The Neck module is used to perform feature fusion on the feature vector of the front side of the sample, the feature vector of the side of the sample and the feature vector of the back of the sample to obtain a feature map of the front side of the sample, a feature map of the side of the sample and a feature map of the back of the sample;
[0129] The Head module is used to predict and classify the front feature image, side feature image and back feature image of the sample clothing, and judge whether the front image, side image and back image of the sample clothing are standardized according to the confidence level.
[0130] In one embodiment, judging whether the front view, side view and back view of the sample garment meet the specification according to the confidence level includes:
[0131] If it is determined according to the confidence level that the sample clothing front image is the front of the sample clothing and the human body / mannequin in the sample clothing front image has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing front image meets the specification;
[0132] If it is determined according to the confidence level that the sample clothing side view is a sample clothing side view and the human body / mannequin in the sample clothing side view has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing side view meets the specification;
[0133] When it is determined based on the confidence level that the sample back image is the back of the sample and the human body / mannequin in the sample back image has an upper body, a lower body, arms, shoulders and a whole body, it is determined that the sample back image meets the specifications.
[0134] In one embodiment, it further includes:
[0135] After the detection result of the sample front image, the sample side image or the sample back image is that the number of times it does not meet the specifications exceeds a third preset threshold, the sample front image, the sample side image or the sample back image is uploaded.
[0136] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related 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 may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of 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 DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.
[0137] 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.
Claims
1. A method for detecting the standardization of a sample clothing image, characterized in that: include: Obtain a training sample set, a front image of a sample garment, a side image of a sample garment, and a back image of a sample garment; The training sample set is used to train the neural network to obtain a sample clothing standard detection model; Inputting the sample clothing front view, sample clothing side view and sample clothing back view into the sample clothing detection standard model to obtain the detection result; Determine whether to upload the front view, side view and back view of the sample garment according to the detection result.
2. The method for detecting the standardization of sample clothing images according to claim 1, characterized in that: The training sample set includes: overall clothing sample images, partial clothing sample images and non-clothing sample images; the overall clothing sample images are positive samples; the partial clothing sample images and non-clothing sample images are negative samples.
3. The method for detecting the standardization of sample clothing images according to claim 2, characterized in that: The method of using the training sample set to train a neural network to obtain a sample clothing standard detection model includes: Pre-training the neural network using the positive samples and negative samples; Dividing the positive samples into a first test set and a first scene set, and dividing the negative samples into a second test set and a second scene set; Fine-tune the neural network using the first test set, the second test set, and the trajectory generated by the pre-trained neural network until the training result reaches a first preset threshold; The neural network is fine-tuned using the first scene set, the second scene set, and the trajectory generated by the fine-tuned neural network until the training result reaches a second preset threshold, thereby obtaining a sample clothing specification detection model.
4. The method for detecting the standardization of clothing sample images according to claim 3, characterized in that: The using the positive samples and the negative samples to pre-train the neural network includes: Adding corresponding sample labels to the positive samples and negative samples; The positive samples and negative samples are used to train the neural network to obtain corresponding training detection categories, a loss function is calculated according to the training detection categories and corresponding sample labels, and the neural network parameters are updated.
5. The method for detecting the standardization of sample clothing images according to claim 1, characterized in that: The neural network is a YOLO neural network; the sample clothing standard detection model includes: an input module, a Backbone module, a Neck module and a Head module; the sample clothing front view, the sample clothing side view and the sample clothing back view are input into the sample clothing detection standard model to obtain a detection result, including: Inputting the sample garment front view, the sample garment side view and the sample garment back view to an input module, wherein the input module is used to scale and crop the sample garment front view, the sample garment side view and the sample garment back view; The Backbone module is used to extract features from the scaled and cropped front image, side image and back image of the sample garment to obtain a sample garment front feature vector, a sample garment side feature vector and a sample garment back feature vector; The Neck module is used to perform feature fusion on the feature vector of the front side of the sample, the feature vector of the side of the sample and the feature vector of the back of the sample to obtain a feature map of the front side of the sample, a feature map of the side of the sample and a feature map of the back of the sample; The Head module is used to predict and classify the front feature image, side feature image and back feature image of the sample clothing, and judge whether the front image, side image and back image of the sample clothing are standardized according to the confidence level.
6. The method for detecting the standardization of clothing sample images according to claim 5, characterized in that: The judging, based on the confidence level, whether the front view, side view and back view of the sample garment meet the specifications includes: If it is determined according to the confidence level that the sample clothing front image is the front of the sample clothing and the human body / mannequin in the sample clothing front image has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing front image meets the specification; If it is determined according to the confidence level that the sample clothing side view is a sample clothing side view and the human body / mannequin in the sample clothing side view has an upper body, a lower body, an arm, a shoulder and a whole body, then it is determined that the sample clothing side view meets the specification; When it is determined based on the confidence level that the sample back image is the back of the sample and the human body / mannequin in the sample back image has an upper body, a lower body, arms, shoulders and a whole body, it is determined that the sample back image meets the specifications.
7. The method for detecting the standardization of clothing sample images according to claim 1, characterized in that: Also includes: After the detection result of the sample front image, the sample side image or the sample back image is that the number of times it does not meet the specifications exceeds a third preset threshold, the sample front image, the sample side image or the sample back image is uploaded.
8. A sample clothing standardization detection system, characterized in that: include: An acquisition module is used to acquire a training sample set, a front view of a sample garment, a side view of a sample garment, and a back view of a sample garment; A training module, used to train a neural network using the training sample set to obtain a sample clothing standard detection model; A detection module, used for inputting the sample clothing front view, the sample clothing side view and the sample clothing back view into the sample clothing detection standard model to obtain a detection result; The judgment module is used to judge whether to upload the front view, side view and back view of the sample clothing according to the detection result.
9. 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.
10. 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.
