Manta Ray Foraging Optimizer with Deep Learning Assisted Automated Fabric Defect Detection

Detecting defects in fabric is a crucial step in the textile industry. The intricate nature of textile structures often poses a challenge for the automated identification of fabric damage. Fabric Defect Detection (FDD) relies on sophisticated methods such as computer vision and machine learning to automatically identify and categorize faults or irregularities in fabric materials. By employing high-resolution imaging methods and classy models, these systems analyze textures, patterns, and colour differences in textiles to identify faults such as holes, stains, or abnormalities. FDD utilizing Deep Learning (DL) influences the power of neural networks to spontaneously recognize and label deficiencies in fabric materials. Through Convolutional Neural Networks (CNNs) or other DL architectures, these techniques are trained on great image datasets that contain defective as well as non-defective textile samples. This research develops a novel Manta Ray Foraging Optimizer with a Deep Learning-Assisted Automated Fabric Defect Detection (MRFODL-AFDD) approach. The MRFODL-AFDD approach uses an Inception v3 feature extractor to improve the efficacy of feature representation in textile images. Besides, the presented approach combines MRFO model for hyperparameter tuning, enhancing the performance of the defect detection method. At last, a deep Long Short-Term Memory (LSTM) detection model is implemented to diagnose temporal needs in consecutive fabric data. Experimental results determine the efficiency of the developed framework in attaining precise and robust FDD. The integration of Inception v3, MRFO, and deep LSTM classification displays a promising improvement in mechanizing excellence control procedures within the textile business.