Fabric Defect Detection Using EfficientNetB0 and Inception V3: A Deep Learning Approach

In any industry, the quality of the final finished product is very crucial. The textile industry is an industry where there is a large variety of final finished products. Common defects include holes and cuts, thread defects, stains, and the presence of foreign bodies. Many industries still use traditional methods of manual inspection. Manual inspection is time-consuming and may have chances of human error, which leads to loss and wastage of resources. To address this challenge, this research paper proposes an automated method of fabric defect detection and classification. To enhance the accuracy and efficiency of fabric defect detection a deep learning approach based on EfficientNetB0 and Inception V3 architectures is proposed.The proposed method leverages the strengths of EfficientNetB0 and Inception V3 and combines their capabilities to detect defects in the fabrics. The developed model is trained and tested using the TILDA database. Experimental results show that the achieved average accuracy is 96.25% for type C1 and 92.5% for type C2 fabrics. The study's findings demonstrate the potential benefits of incorporating cutting-edge deep learning models into fabric inspection systems.