| ABSTRACT |
This research paper introduces an automated system for fabric defect detection, crucial for maintaining quality in the textile industry. Traditional methods of inspection are often manual, time-consuming, and prone to human error, leading to inconsistent quality control. To address these challenges, we have developed a deep learning-based approach utilizing a Convolutional Neural Network (CNN) to classify various fabric defects. The system takes an image of fabric as input, processes it using a pre-trained MobileNetV2 architecture fine-tuned on a custom dataset of fabric images, and outputs the identified defect type along with a confidence score. To enhance interpretability and provide visual evidence of the model's decision-making process, we have integrated Grad-CAM (Gradient-weighted Class Activation Mapping) to generate heatmaps highlighting the regions of interest in the fabric image that led to the defect classification. Furthermore, the system provides a user-friendly web interface built with Flask, allowing for easy upload of fabric images and visualization of results. A quality rating based on the confidence of the prediction is also provided, along with actionable suggestions for defect mitigation. |
