Smart Fabric Inspection: Leveraging Deep Learning for Defect Detection

The early identification of defects on fabrics is important especially when it comes to the production of textile products. It presents the development of the deep learning technique for automated detection of defect zones in fabric. Using the sophisticated architectural design of Convolutional Neural Network (CNN), Mobile Net, Dense Net and Alex Net the research seeks to improve accuracy of the defects. In particular, an autoencoder model is applied due to its data compression and decompression features, which enable feature extraction. Thus, the study is cantered on the adjustment of the autoencoder’s input values, as well as hyperparameters for enhancing the performance. Holistically deploying the CNN in combination with other compact architectures such as the Mobile Net in combination with sound architectures such as Dense Net and Alex Net offers a versatile solution for fabric defect detection. This research proves by using different types of deep learning models combined, the automatic fault detection system in textile manufacturing can be enhanced in both accuracy and efficiency.