Abstract
Fabric defects impact textile quality and production efficiency. This study proposes an Optimized Attention Augmented Residual Convolutional Neural Network with FA-ResNet for Fabric Defect Detection (AARCNN-FA-ResNet-FDD). Images from a fabric defect dataset are pre-processed using Multiple Local Particle Filter (MLPF), followed by feature extraction using Synchro Transient Extracting Transform (STET). Key texture features are classified using AARCNN-FA-ResNet, with parameters optimized via the Lotus Effect Optimization Algorithm (LEOA). The method significantly improves classification accuracy, precision, and reduces computation time compared to existing models like FSDC-CSO-DRN, LSTM-TC-FDD, and CNN-ATCD-FDD, achieving up to 99% classification accuracy.
DOI: 10.61416/ceai.v27i4.9588