A Fabric Defect Detection Method Based on Deep Learning

Abstract

Fabric defect detection is a challenging task in the fabric industry because of the complex shapes and large variety of fabric defects. Many methods have been proposed to solve this problem, but their detection speed and accuracy were very low. As a classic deep learning method and end-to-end target detection algorithm, YOLOv4 has evolved rapidly and has been applied in many industries, showing good performance. This paper proposes an improved YOLOv4 algorithm with higher accuracy for fabric defect detection, in which a new SPP structure that uses SoftPool instead of MaxPool is adopted. The improved YOLOv4 algorithm with three SoftPools can process the feature map effectively, which has a significant advantage in reducing the negative side effects of the SPP structure and improving the detection accuracy. The improved SPP structure is used by the three outputs of Backbone, and in order to ensure that the output can be inputted into the subsequent PANet successfully, the network structure is improved that a series of convolution layers after the SPP structure is added for reducing the channel numbers of feature map to an appropriate value. In addition, contrast-limited adaptive histogram equalization is adopted in advance to improve the image quality, which results in strong anti-interference abilities and can slightly increase the mAP. Experimental results show that, compared with the original YOLOV4, the improved YOLOv4 increases the mAP effectively by 6%, while the FPS only decreases by 2. The improved YOLOv4 can identify the location of defects accurately and quickly, and can also be applied in other defect detection industries.

A Lightweight Detector Based on Attention Mechanism for Fabric Defect Detection

“…As observed from the table, the present experimental results were 5.62% more accurate than those reported by the first-place winner of the Tianchi industrial competition. For validation, we compared the current experimental results with those obtained by Liu et al [23,24,32], who applied the fabric defect dataset of Tianchi. As observed, the present model displayed adequate results for individual defects, and the detection accuracy of large defects such as stain, hundred feet, and other difficult-to-detect defects such as abrasion mark were optimal compared to other studies.…”

“…In fabric defects such as stains and grains that account for a small fabric area and are difficult to detect with the naked eye, the application of YOLOv4 cannot be migrated to the field of fabric defect detection. In this regard, Liu et al [32] proposed a fabric defect detection method based on YOLOv4 with high detection accuracy, but it divided the fabric defect dataset only into four categories, all of which included easily detectable fabric defect types such as lines and holes, and no targeted research based on minor defects was conducted.…”

“…The YOLO Head was used for the module to determine whether the fabric defects are identifiable and the type of identified defects corresponds to the three priori frames preset in the feature enhancement networks. In particular, the three tensor dimensions of the input feature-(64, 64, 256), (32,32,512), and (16, 16, 1024)-with output dimensions (after Head) of (64 64, 21), (32,32,21), (16,16,21), and 21 in the last dimension contains the x_offset, y_offset, image width and height, confidence level, and defect classification result. The decoding process of the head first involves adding each grid point with its corresponding x_offset and y_offset to derive the center of the prediction frame.…”

Research on improving YOLOv5s algorithm for fabric defect detection

“…Kumar (2008) applied YOLOv2 to fabric defect detection and attained a fabric defect detection accuracy of 98%, but their missed detection rate for small targets is relatively high. Liu et al (2022) adopted a new SPP structure using SoftPool instead of MaxPool and integrated it into the YOLOv4 algorithm. They effectively improved the average accuracy but had a high missed detection rate for small target defects.…”

Flaw Detection in White Porcelain Wine Bottles Based on Improved YOLOv4 Algorithm

“…Res2Net was used as the backbone feature extraction network, resulting in average detection accuracy of the algorithm up to 80.1%, and the detection speed up to 76.9 frames/s. Liu Q. et al (2022 ) proposed an improved YOLOv4 algorithm for fabric flaw detection. A new SPP structure was adopted and SoftPool was used instead of MaxPool, which made the average accuracy of fabric flaw detection achieve 86.5%.…”