Novel Multipath Convolutional Neural Network Based Fabric Defect Detection System

Detecting defects in fabric is one of the most important steps in the process of quality control in manufacturing processes. The textile structure can deviate from the design due to improper mechanical motion or yarn breakage on a loom, producing a warp, weft, or point defect like harness misdraw, endout, mispick, and slub. Visual human inspection results in common mistakes and takes more time, both of which might reduce productivity. Therefore, automated fabric defect identification will save time and enable more accurate and rapid defect prediction. Due to the Convolution Neural Network's high level of image classification and recognition accuracy, it is utilised to detect fabric defects. It chooses just appropriate features for object identification from a vast number of created features. The proposed model works on the multipath CNN concept, where first path is CNN with tanh activation layer + GLCM and the second path is VGG – 16 + Gabor. The novel multipath CNN was evaluated using TILDA dataset with total of 2000 images and simulated for 20 epochs.