An automatic dyeing method based on improved YOLOv11 instance segmentation network for visual recognition

Abstract

The present application relates to the technical field of clothing intelligent printing and dyeing and textile automatic processing, and particularly relates to an automatic injection dyeing method based on improved YOLOv11 instance segmentation network for visual identification, comprising the following steps: S1: shooting the finished product clothes after injection dyeing, so as to obtain a training data set; S2: marking the training set, training and building an improved YOLOv11 network model; S3: placing the sample not subjected to injection dyeing on a material frame, and manually setting the injection dyeing color required by the robot according to the injection dyeing effect; S4: shooting the sample clothes placed; S5: after shooting, using the trained network model to perform visual identification on the obtained pictures, and obtaining the Mask mask of the clothes and the marker in the pictures; S6: using Opencv traditional image processing. The present application can use the improved neural network model to perform visual identification on the injection dyed clothes, provide positioning for subsequent robot automatic injection dyeing, greatly improve the production efficiency, and reduce the labor operation cost.

Description

Technical Field

[0001] This invention relates to the field of intelligent dyeing and printing technology for clothing and automated processing technology for textiles, and in particular to an automated dyeing method based on visual recognition using an improved YOLOv11 instance segmentation network. Background Technology

[0002] In recent years, with the continuous development of automation and intelligent manufacturing technologies, traditional garment dyeing processes have become unable to meet the ever-increasing demands for production efficiency and quality. To improve production efficiency and reduce labor costs, more and more garment manufacturers and dyeing factories are beginning to introduce robotics and computer vision technologies to automate the dyeing process. However, although robotics can simulate human operation, it still faces many technical challenges during execution, especially in the accurate recognition and positioning of garment outlines.

[0003] In traditional dyeing processes, especially the coloring stage, manual labor is heavily relied upon, requiring 6-8 workers at almost every station. Furthermore, the dyeing process is susceptible to the effects of worker skill and fatigue. Therefore, achieving accurate visual recognition of garments using computer vision and artificial intelligence has become a core technology for automated dyeing systems. Due to the irregular shape of garments during dyeing, the penetration of pigments, color gradations, and the fact that the dyeing process takes place on a transparent mesh frame (allowing for excess pigment penetration), all these factors affect visual recognition. The key to the successful application of this technology lies in achieving high-precision garment outline recognition and marker localization under complex textures and background interference, as well as selecting a suitable network model for visual recognition.

[0004] The existing YOLOv11 (You Only Look Once) deep learning model has become one of the mainstream methods in the field of visual recognition due to its efficiency and accuracy in object detection and instance segmentation tasks. Especially in instance segmentation tasks, YOLOv11 can accurately segment objects and identify the contours of irregularly shaped objects. However, the existing YOLOv11 instance segmentation model still has certain limitations in dealing with complex backgrounds, texture transitions, and color interference, especially in the accurate identification of clothing contours and small target markers during the dyeing process, where there is still room for improvement.

[0005] Patent CN113684633B proposes an automatic injection dyeing machine and also mentions related injection dyeing sampling technology. It achieves garment positioning and dyeing through mechanical positioning and fixing components. However, injection-dyed fabrics have significant shape variations and complex contours, making precise positioning difficult to achieve mechanically. Patent CN115710787B proposes a human-machine collaborative intelligent spraying robot; however, although this equipment partially automates the dyeing operation, its operation still relies on manual intervention. Operators need to continuously monitor nozzle blockage during production and manually clean and adjust the equipment's operation when necessary.

[0006] To overcome the aforementioned technical limitations, this invention achieves intelligent and automated production of the entire process by automatically adjusting the spraying path and dye spraying strategy through visual recognition, making it suitable for mass production of complex textures and personalized patterns. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing an automated dyeing method based on visual recognition using an improved YOLOv11 instance segmentation network. By introducing the YOLOv11 instance segmentation network into the automated dyeing system, the dye spraying and positioning operations are dynamically controlled through real-time identification of clothing outlines and marker areas, significantly improving the automation level and dyeing accuracy of the dyeing process.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] An automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network includes the following steps:

[0010] Step 1: Prepare training dataset

[0011] Set up the shooting environment and field of view, place the dyed finished clothes on the material frame and take pictures to obtain multiple training datasets;

[0012] Step 2: Construct a neural network

[0013] The clothing outlines and landmarks in the training dataset are labeled, and the improved YOLOv11 neural network model is built by training on these two features after labeling.

[0014] Step 3: Manually set the robot program

[0015] After the model is trained, undyed clothing samples are placed on the material frame, and the robot's execution program is set according to the desired dyeing effect, such as the number and types of colors to be dyed on the clothes.

[0016] After setting the program, save the program, and the robot can then specify the dyeing color for different areas according to the program.

[0017] Step 4: Take pictures of the sample clothing.

[0018] Take photos of the displayed sample garments;

[0019] Step 5: Visual Recognition

[0020] After the photos are taken, the trained network model is used to perform visual recognition on the captured images to obtain the mask codes for clothing and landmarks in the images.

[0021] Step 6: Image Processing

[0022] Using OpenCV's traditional image processing, the recognized image is processed. First, the outer contour point set of the clothing and marker mask is obtained and connected to obtain the contour. Then, the marker line is obtained by connecting the center of the circle, the marker and the outer contour of the clothing based on the center of the clothing and the center of the marker.

[0023] Step 7: Robotic processing and dyeing

[0024] The robot performs the dyeing process, dyeing within the outline of the garment and changing the dye color at the marked lines.

[0025] Preferably, in step 1, the specific steps are as follows:

[0026] Step 1.1: The shooting environment should minimize the impact of ambient light changes on subsequent visual recognition. Therefore, two strip light sources were added, and their brightness and the camera's exposure parameters were adjusted to meet the requirements.

[0027] Step 1.2: The dataset captured should minimize the impact of the background on subsequent recognition and ensure uniformity of the field of view. Therefore, adjust the field of view to just fill the entire material frame.

[0028] In the training set, the two features of clothing and markers are used to identify the mask of clothing and markers in the dyeing sample making task. The color changes of dye and the complexity of clothing outlines, as well as the fact that the dyeing pigment can easily seep into the material frame due to the permeable mesh frame, can affect the recognition effect. Therefore, it is very challenging to achieve the recognition accuracy.

[0029] Preferably, in step 2, the specific steps of the improved YOLOv11 neural network model are as follows:

[0030] Embedding the CBAM (Convolutional Block Attention Module) into the feature fusion stage of YOLOv11 is beneficial because the main task of the Neck part in the YOLOv11 network architecture is to fuse feature maps from different resolutions. Therefore, introducing CBAM into the C3k2 module can enhance the model's ability to perceive details (such as clothing edges and colors), enabling it to handle these difficult-to-distinguish details more effectively and reduce the interference of the background frame. It can also help the network focus on important features at different scales, thereby improving the recognition accuracy of the head-large detector for clothing outlines and the head-small detector for landmarks.

[0031] The CBAM module is essentially an attention module integrated into the convolutional neural network, designed to enhance the model's feature representation capabilities. The CBAM module consists of a Channel Attention Module and a Spatial Attention Module, which are applied to two dimensions of the feature map: the channel dimension and the spatial dimension, respectively, to enhance the features in both dimensions.

[0032] Preferably, the channel attention module includes the following steps:

[0033] a. Perform global average pooling and global max pooling on the feature map to obtain a global descriptor;

[0034] b. Perform dimensionality reduction and activation (usually ReLU activation) on the descriptor through a fully connected layer;

[0035] c. Finally, the weights of each channel are obtained through sigmoid activation, and these weights are used to weight the channels of the feature map;

[0036] The spatial attention module includes the following steps:

[0037] a. Perform average pooling and max pooling on the feature map along the channel dimension to obtain two single-channel feature maps;

[0038] b. After concatenating these two feature maps, pass them through a convolutional layer to obtain the spatial attention weight map;

[0039] c. Finally, the spatial weights are restricted to the [0,1] interval by the sigmoid activation function, and each position of the feature map is weighted according to this weight.

[0040] Preferably, after establishing the spatial attention module and the channel attention module, CBAM concatenates and mixes the channel attention module and the spatial attention module together to perform two enhancement processes on the feature map. The specific steps are as follows:

[0041] a. Channel attention enhancement: First, the input feature map F passes through the channel attention module to generate attention weights and perform weighted processing;

[0042] b. Spatial attention enhancement: Next, the feature map after channel enhancement is fed into the spatial attention module, where spatial attention weights are generated and weighted to obtain the final output feature map.

[0043] This section introduces the CBAM module. By incorporating these two attention mechanisms, CBAM helps YOLOv11 focus on important features such as the outlines of clothing and markers, and dye boundaries in complex scenes, avoiding interference from background or irrelevant areas. Through spatial attention, CBAM dynamically adjusts its focus at different spatial locations. Since the clothing to be dyed is all on the fabric frame, the spatial attention module enhances features in important areas while suppressing the influence of irrelevant areas, ensuring the model focuses on key areas within the frame, thereby improving segmentation accuracy and reducing interference from irrelevant information. It's worth noting that the CBAM module is relatively lightweight and does not significantly increase computational overhead compared to other attention mechanisms, making it suitable for use in conjunction with YOLOv11's efficient detection and segmentation architecture.

[0044] Subsequently, the improved YOLOv11 instance segmentation network model was trained, and the training results were exported as .onnx files and deployed into the visual recognition program.

[0045] Preferably, step 6 includes the following steps:

[0046] a. First, obtain the outer contour point set of the clothing and logo mask, and then connect the outer contour point set to obtain the outer contour of the clothing;

[0047] b. Obtain the center of the clothing outline O1 and the centers of other marker outlines O2, O3…O n ;

[0048] c. Starting from O1, along the path from the center point O of the marker... i Find the intersection point O of the line connecting (i = 2, 3, ..., n) with the outline of the clothing. i ', connect O1 O i 'That is, the marker line.'

[0049] Preferably, step 7 includes the following steps:

[0050] a. After the robot obtains the position coordinates of the clothing outline, the program is set to process and color it inside the clothing outline;

[0051] b. Based on the quantity and location information of the markers, the program sets their processing areas, and different processing areas are dyed with different colors according to the colors set manually.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] This invention is used to accurately identify garment outlines and markers during the dyeing and sampling stage. Through improvements to the YOLOv11 model, it maintains high-precision recognition performance even when facing challenges such as complex dyeing textures, color gradients, and pigment penetration. This improvement not only enables successful identification of the outlines and markers of different garments but also provides precise positioning and control for the robot in subsequent automated dyeing processes. Ultimately, it achieves complete automation of the dyeing task, significantly improving production efficiency, reducing manual operation costs, and enhancing quality control capabilities throughout the dyeing process. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the garment sample before dyeing in this invention;

[0055] Figure 2 This is a schematic diagram showing the details of the material frame in this invention;

[0056] Figure 3 This is a diagram of the improved YOLOv11 network structure in this invention;

[0057] Figure 4 This is a schematic diagram of the channel attention module in this invention.

[0058] Figure 5 This is a schematic diagram of the spatial attention module in this invention.

[0059] Figure 6 This is a schematic diagram of the CBAM hybrid attention module in this invention.

[0060] Figure 7 This is a schematic diagram illustrating the processing results of using OpenCV to process the recognized image to obtain the outer contour of the clothing and the marking lines in this invention.

[0061] Figure 8 This is a schematic diagram illustrating the processing effect after robot dyeing in this invention;

[0062] Figure 9 This is a flowchart of the present invention. Detailed Implementation

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings, so that those skilled in the art can better understand the advantages and features of the present invention, thereby making a clearer definition of the scope of protection of the present invention. The embodiments described in this invention are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0064] An automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network includes the following steps:

[0065] (1) Prepare the training dataset;

[0066] First, the shooting environment was set up and the camera's field of view was adjusted. The shooting environment should minimize the impact of ambient light variations on subsequent visual recognition; therefore, an additional light source was added, and its brightness and the camera's exposure parameters were adjusted to meet the requirements. The dataset being shot should minimize the influence of the background on subsequent recognition and ensure a uniform field of view; therefore, the field of view was adjusted to exactly fill the entire frame.

[0067] Then, the undyed sample garments were placed on the material frame, and markers were affixed to the color-changing locations and photographed. Figure 1 This refers to circular sample garments that have not been dyed. During the data acquisition process, the garments should be varied in multiple poses as much as possible to improve the robustness of recognition and thus obtain multiple training datasets.

[0068] (2) Construct a neural network;

[0069] After taking the photos, the clothing outlines and markers in the training set images are labeled. After labeling, these two features are analyzed, and the training and validation sets needed for training are compiled. Since the goal of the dyeing sample-making task is to identify the clothing outline and rectangular color markers, these features are influenced by dye color variations, the complexity of the clothing outline, and the fact that the fabric frame is a permeable mesh frame. Figure 2 As shown in the detailed diagram of the mesh frame, the injected pigment can easily seep into the bottom of the frame to some extent, which can affect the recognition effect. Therefore, it is very challenging to achieve the accuracy of the recognition.

[0070] YOLOv11 is the latest generation of the YOLO (You Only Look Once) series of neural networks, further improving the performance of object detection and instance segmentation tasks. As a high-performance real-time deep learning network, YOLOv11 combines efficient feature extraction capabilities with a multi-task learning framework, achieving higher detection and segmentation accuracy through improved network architecture and loss function.

[0071] YOLOv11's instance segmentation capability is based on pixel-level segmentation features introduced in its object detection framework. Its design allows for precise segmentation of each pixel of the target while simultaneously detecting the target's bounding box. This capability enables YOLOv11 to accurately predict the boundaries of targets in complex scenes, while maintaining efficiency and robustness in segmentation tasks.

[0072] Because clothing outline segmentation is quite complex, YOLOv11 still sometimes produces inaccurate results. Therefore, to further improve its performance, some modifications were made to the basic YOLOv11 network architecture. The CBAM (Convolutional Block Attention Module) module was embedded into the feature fusion stage of YOLOv11. Since the main task of the Neck part in the YOLOv11 network architecture is to fuse feature maps from different resolutions, introducing CBAM into the C3k2 module of the Neck part allows the model to have a stronger perception of details (such as clothing edges and colors), more effectively handling these difficult-to-distinguish details and reducing the interference of the background frame. It also helps the network focus on important features at different scales, thereby improving the recognition accuracy of the head-large head for clothing outlines and the head-small head for landmarks. The improved network architecture is as follows: Figure 3 As shown.

[0073] The CBAM module is essentially an attention module integrated into the convolutional neural network, designed to enhance the model's feature representation capabilities. It consists of a ChannelAttentionModule and a SpatialAttentionModule. These modules can be applied to the two dimensions of the feature map: the channel dimension and the spatial dimension, respectively, to enhance the features in both dimensions. The following is an analysis of these two modules in CBAM:

[0074] ①Channel attention mechanism:

[0075] like Figure 4 The schematic diagram of the channel attention module is shown below. The processing steps are as follows:

[0076] a. For the input feature map F∈R C×H×W (Where C is the number of channels, and H and W represent the height and width, respectively), first, global max pooling and global average pooling operations are performed on each channel to obtain the global description F. avg ∈R C×1×1 and F max ∈R C×1×1 :

[0077]

[0078] In the formula, F avg (c) represents the result of average pooling for the c-th channel, F max (c) represents the result of average pooling for the c-th channel.

[0079] b. After extracting feature vectors through global max pooling and global average pooling operations, these two types of features are fed into a shared fully connected layer. This fully connected layer is used to learn the attention weights for each channel, allowing the network to adaptively determine which channels are more important to the current task. Finally, the final attention weight vector is generated by combining the feature vectors from global max pooling and average pooling.

[0080] MLP(x) = W1·ReLU(W0·x)

[0081] In the formula, W0 and W1 are learnable weight matrices, and ReLU is a non-linear activation function.

[0082] c. To constrain the attention weights to the range of 0 to 1, a sigmoid activation function is applied to generate attention weights for each channel. These weights are then applied to each channel of the original feature map.

[0083] M c =σ(MLP(F) avg )+MLP(F max ))

[0084] In the formula, σ is the sigmoid function, which ensures that the output is between 0 and 1.

[0085] d. Multiply the generated attention weights by each channel of the original feature map to obtain a weighted channel feature map. This operation can highlight channels that contribute to the current task while suppressing channels that are irrelevant to the task.

[0086] F′=M c ⊙F

[0087] Where ☉ represents element-wise multiplication.

[0088] ② Spatial attention mechanism:

[0089] like Figure 5 The schematic diagram of the spatial attention module is shown below. The processing steps are as follows:

[0090] a. For the input feature map F∈R C×H×W(Where C is the number of channels, and H and W represent the height and width, respectively), first, global max pooling and global average pooling operations are performed on each channel to obtain the global description. and

[0091]

[0092] In the formula This represents the channel average at position (i, j). This represents the maximum value of the channel at position (i, j).

[0093] b. Concatenate the features obtained from max pooling and average pooling along the channel dimension to form a feature map that integrates contextual information from different scales.

[0094]

[0095] In the formula, [;] represents the splicing operation on the channel dimension.

[0096] c. Next, the feature map is convolved using a 7×7 convolution kernel to generate the corresponding spatial attention weights. Then, the spatial weights are restricted to the [0,1] interval using a sigmoid activation function.

[0097]

[0098] In the formula, σ is the sigmoid function, and f 7×7 It is a 7×7 convolution operation.

[0099] d. Apply the generated spatial attention weights to the original feature map to weight the features at each spatial location, thereby highlighting key image regions while reducing the influence of unimportant regions.

[0100] F″=M S ⊙F

[0101] With the spatial attention module and the channel attention module, CBAM concatenates and mixes the channel attention module and the spatial attention module together, performing two enhancement processes on the feature map. The schematic diagram is as follows: Figure 6 As shown, the entire process is as follows:

[0102] a. First, the input feature map F passes through the channel attention module to generate attention weights M. c The enhanced feature map F is obtained by weighting the features. c .

[0103] b. Next, the feature map F after channel enhancement c The input is passed to the spatial attention module, which generates spatial attention weights M.s The feature map F is then weighted to obtain the final output feature map. final The entire processing flow of the CBAM module can be represented as follows:

[0104] F final =M s (M c (F)⊙F)⊙F

[0105] The introduction of the CBAM module, which incorporates both attention mechanisms, allows YOLOv11 to focus on key features such as the outlines of clothing and markers, and dye boundaries in complex scenes through channel attention, avoiding interference from background or irrelevant areas. Through spatial attention, CBAM dynamically adjusts its focus at different spatial locations. Since the clothing to be dyed is located on the fabric frame, the spatial attention module enhances features in important areas while suppressing the influence of irrelevant areas, ensuring the model focuses on key regions within the frame, thereby improving segmentation accuracy and reducing interference from irrelevant information. Notably, the CBAM module is relatively lightweight and does not significantly increase computational overhead compared to other attention mechanisms, making it suitable for integration with YOLOv11's efficient detection and segmentation architecture.

[0106] Subsequently, the improved YOLOv11 instance segmentation network model was trained, and the training results were exported as .onnx files and deployed into the visual recognition program.

[0107] (3) Manually setting the robot program;

[0108] After the model is trained, undyed clothing samples are placed on the material frame, and the robot's execution program is set manually according to the desired dyeing effect, such as the number and types of colors to be dyed on the clothing.

[0109] After setting the program, save the program, and the robot can then specify the dyeing color for different areas according to the program.

[0110] (4) Take pictures of the sample clothing;

[0111] Take photos of the displayed sample clothes.

[0112] (5) Visual recognition;

[0113] An improved YOLOv11 instance segmentation network model was used to identify images taken after teaching, and the final identification results were mask masks of clothing outlines and landmark outlines.

[0114] (6) Image Processing

[0115] Using OpenCV's traditional image processing techniques, the recognized images are processed as follows:

[0116] a. First, obtain the outer contour point set of the clothing and marker mask, and then connect the outer contour point set to obtain the outer contour of the clothing.

[0117] b. Obtain the center of the clothing outline O1 and the centers of other landmark outlines O2, O3…O n .

[0118] c. Starting from O1, along the path from the center point O of the marker... i Find the intersection point O of the line connecting (i = 2, 3, ..., n) with the outline of the clothing. i ′, connecting O1O i ′ is the marker line.

[0119] The processed clothing outline and markings are as follows: Figure 7 As shown.

[0120] (7) Robotic processing and dyeing

[0121] The robot performs the dyeing process, applying dye within the garment's outline and changing the dye color at the marked lines. The final effect is as follows. Figure 8 As shown, robots can replace manual labor in the automated dyeing process, achieving the same expected results while reducing manual operation costs.

[0122] In summary, this invention is used to accurately identify garment outlines and markers during the dyeing and sampling stage. Through improvements to the YOLOv11 model, it maintains high-precision recognition performance even when facing challenges such as complex dyeing textures, color gradations, and pigment penetration. This improvement not only enables successful identification of the outlines and markers of different garments but also provides precise positioning and control for the robot in subsequent automated dyeing processes, ultimately achieving full automation of the dyeing task. This significantly improves production efficiency, reduces manual operation costs, and enhances quality control capabilities in the dyeing process.

[0123] The descriptions and practices disclosed in this invention are readily apparent and understandable to those skilled in the art, and various modifications and refinements can be made without departing from the principles of this invention. Therefore, any modifications or improvements made without departing from the spirit of this invention should also be considered within the scope of protection of this invention.

Claims

1. An automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network, characterized in that, Includes the following steps: Step 1: Prepare training dataset Set up the shooting environment and field of view, place the dyed finished clothes on the material frame and take pictures to obtain multiple training datasets; Step 2: Construct a neural network The clothing outlines and landmarks in the training dataset are labeled, and the improved YOLOv11 neural network model is built by training on these two features after labeling. Step 3: Manually set the robot program The undyed sample is placed on the material frame, and the operator sets the dyeing color required by the robot based on the dyeing effect. Step 4: Take pictures of the sample clothing. Take photos of the displayed sample garments; Step 5: Visual Recognition After the photos are taken, the trained network model is used to perform visual recognition on the captured images to obtain the mask codes for clothing and landmarks in the images. Step 6: Image Processing Using OpenCV's traditional image processing, the recognized image is processed to obtain the clothing outline and the position of the marking lines; Step 7: Robotic processing and dyeing The robot performs the dyeing process, dyeing within the outline of the garment and changing the dye color at the marked lines.

2. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 1, characterized in that, In step 1, the specific steps are as follows: Step 1.1: The shooting environment should be designed to minimize the impact of ambient light changes on subsequent visual recognition. Add two strip light sources and adjust their brightness and the camera's exposure parameters to meet the requirements. Step 1.2: The captured dataset should avoid the background affecting subsequent recognition and ensure uniformity of the field of view. Adjust the field of view to fill the entire material frame.

3. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 1, characterized in that, In step 2, the specific steps of the improved YOLOv11 neural network model are as follows: The CBAM module is embedded into the feature fusion stage of YOLOv11. The CBAM module is an attention module integrated into the convolutional neural network, which aims to enhance the feature representation capability of the model. The CBAM module consists of a channel attention module and a spatial attention module. These are applied to two dimensions of the feature map: the channel dimension and the spatial dimension, respectively, and the features are enhanced in both the channel dimension and the spatial dimension.

4. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 3, characterized in that, The channel attention module includes the following steps: a. Perform global average pooling and global max pooling on the feature map to obtain a global descriptor; b. Reduce and activate the descriptors using a fully connected layer; c. Finally, the weights of each channel are obtained through sigmoid activation, and these weights are used to weight the channels of the feature map; The spatial attention module includes the following steps: a. Perform average pooling and max pooling on the feature map along the channel dimension to obtain two single-channel feature maps; b. After concatenating these two feature maps, pass them through a convolutional layer to obtain the spatial attention weight map; c. Finally, the spatial weights are restricted to the [0,1] interval by the sigmoid activation function, and each position of the feature map is weighted according to this weight.

5. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 3, characterized in that, The specific steps for enhancing features in the channel and spatial dimensions by introducing channel attention and spatial attention modules are as follows: a. Channel attention enhancement: First, the input feature map F passes through the channel attention module to generate attention weights and perform weighted processing; b. Spatial attention enhancement: Next, the feature map after channel enhancement is fed into the spatial attention module, where spatial attention weights are generated and weighted to obtain the final output feature map.

6. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 1, characterized in that, Step 6 includes the following steps: a. First, obtain the outer contour point set of the clothing and logo mask, and then connect the outer contour point set to obtain the outer contour of the clothing; b. Obtain the center of the clothing outline O1 and the centers of other marker outlines O2, O3…O n ; c. Starting from O1, along the path from the center point O of the marker... i Find the intersection point O of the line connecting (i = 2, 3, ..., n) with the outline of the clothing. i ', Connecting to O1O i 'That is, the marker line.' 7. The automated annotation method for visual recognition based on an improved YOLOv11 instance segmentation network according to claim 1, characterized in that, Step 7 includes the following steps: a. After the robot obtains the position coordinates of the clothing outline, the program is set to process and color it inside the clothing outline; b. Based on the quantity and location information of the markers, the program sets their processing areas, and different processing areas are dyed with different colors according to the colors set manually.

Images