Fabric Fault and Extra Thread Detection using Convolutional Neural Network

Abstract

A planar substance made of textile fibers is called fabric. The main reason why defective fabrics are produced is loom malfunctions. A specialized computer vision system called a fabric inspection system is used to find fabric flaws in order to ensure product quality. In this paper we classify the defect by using Convolutional Neural Network. Utilizing a special type of class-based ensemble convolutional neural network architecture, the defect recognition system is built. The experiment is carried out using several textile fiber kinds. There is four layers in CNN to classify the defect that is Convolution, ReLU, Pooling, Fully Connected layer. A number of well-known CNN architectures, such as Inception, ResNet, VGG, MobileNet, DenseNet, and Xception to classify the defect are tested in this study. Finally, The study demonstrates the result by classification and proves how accurately the defects are identified.