Fabric Defect Detection Based on Cascade Faster R-CNN
Zhiyong Zhao, Kang Gui, Peimao Wang
Abstract
With the development of the textile production industry, quality inspection has become an increasingly important means of ensuring the quality of textiles. In order to solve the problem of low efficiency of traditional manual detection methods, automatic textile defect detection technology has become a research focus. Deep learning has several prominent advantages for fabric defect detection. However, the existence of various types of fabric defects and the imbalance of categories have brought great challenges to fabric defect detection. Therefore, this paper proposes a cascaded Faster R-CNN network, which classifies different kinds of defective fabrics by a pre-classifier network, and then sends it to the Faster R-CNN network for defect detection. At the same time, considering the particularity of fabric defects, the optimization of NMS (Non-Maximum Suppression) was carried out. Experimental results show that the method proposed in this paper significantly improves the detection effect of fabric defects.