Fabric defect detection based on anchor-free network

Abstract

Fabrics play a pivotal role in human life and production, and surface defects can directly affect the quality and value of fabrics. Many methods for fabric defect detection have been proposed, but tiny defects are still difficult to be detected effectively, and the accuracy of defect localization and classification is low. To address these issues, a modified YOLOX network called YOLOX-CATD is proposed, which was supplemented with a coordinate attention module (CAM) and tiny defect detection layer (TDDL) for fast and efficient detection of fabric defects, especially tiny defects. Firstly, the anchor-free network is used as the detection framework to avoid the influence of hyperparameters of the setting anchor. Secondly, a CAM is proposed to enhance the representation of the object of interest in the input feature map and suppress the background regions. Finally, a TDDL is added to introduce high-resolution features to improve the localization accuracy of tiny defects. The experimental results on the Aliyun Tianchi Fabric dataset and NEU-DET demonstrate the superiority and generalization of the modified model. The mean average precision (mAP) of YOLOX-CATD on the fabric defect dataset is improved by 5.67% compared to the original YOLOX, and the detection speed can reach 35-36 frames per second (FPS). This proves that YOLOX-CATD can obtain excellent fabric defect detection performance and meet the urgent need for real-time detection in industrial applications.

A Novel Dataset for Fabric Defect Detection: Bridging Gaps in Anomaly Detection

“…Wang et al used a modified YOLOv3, with a coordinate attention module and a new tiny defect detection layer, culminating in a new anchor-free detector, YOLOX-CATD, which did not require anchor-related hyperparameter tuning [94].…”

Surface defect detection of sawn timbers based on efficient multilevel feature integration

“…Deep learning constructs nonlinear network models by updating the parameters of convolutional neural networks (CNNs) [7]. The feature extracted by CNNs has excellent universality [8], which enables its application in various vision tasks [9][10][11]. For timber surface defect detection, object detection [12] and semantic segmentation [13] are widely used to estimate the classification and the localization of defects.…”

Surface defect detection of strip steel based on GT-CutMix augmentation algorithm and improved DSSD model

“…Liu and Ma [18] proposed a multi-layer attention mechanism which enabled the network to mine defect information effectively, reduce information loss in adjacent feature maps, and improve the accuracy for strip surface defect detection. Wang et al [19] proposed an improved YOLOX network, YOLOx-CATD, which added a coordinate attention module (CAM) and a small defect detection layer (TDDL) to quickly and efficiently detect small defects. Chen et al [20] proposed a new lightweight LF-YOLOv4 model that improved on deep separable convolution by using the squeezeand-excitation (SE) attention module and replaced some common convolutions in the Neck and Head networks of YOLOv4.…”