Fabric Defect Detection Method Based on Improved U-Net

Abstract Computer vision builds a connection between image processing and industrials, bringing modern perception to the automated industrials. At same time, defect detection based on deep learning has played an important role in detection. In this paper, improved convolutional neural network CU-Net for fabric is proposed. method, classical U-Net was improved. On basis of size compression, attention mechanism introduced new compound loss function used training. Using public AITEX data set as test sample, experimental result shows that accuracy recall proposed method are 98.3% 92.7%, respectively. Compared with highest scores other methods, they by 4.8% 2.3%, which improves significantly.