Smart Fabric Inspector System using Deep Learning Algorithm

The detection of fabric defects is a critical task in the textile industry, directly impacting quality control and production efficiency. Existing methods often struggle with high precision, especially when detecting subtle defects such as small holes, missing threads, or texture irregularities. This research addresses these challenges by proposing a novel deep learning-based framework for fabric defect detection that combines multi-scale feature extraction and advanced segmentation techniques. The model leverages synthetic data generation and augmented training datasets to improve its robustness across diverse fabric types, lighting conditions, and defect variations. Furthermore, the research introduces a customized loss function to optimize precision, aiming to minimize false positives while maintaining high recall. Evaluation on real-world textile datasets demonstrates the model’s superior ability to detect and localize fabric defects accurately, making it suitable for real-world industrial applications.