A Hybrid Deep Learning Framework for Fabric Stain Analysis: YOLOv8 based Detection with ResNet-18 Classification

In the textile industry, identifying the fabric flaws and stains is an essential part of quality control. To achieve this intelligent manufacturing in the textile sector, technology for computer vision inspection is essential for fabric stains. The main aim of my project is in order to recognize and classify the various kinds of stains on fabric automatically. To do this, we are utilizing YOLOv8 and ResNet-18.YOLOv8 is a state-of-art for object recognition framework with a real-time performance algorithms. PyTorch framework is also used to classify stains, the YOLOv8 algorithms have been improved and adapted to more effectively identify and classify stains in fabric images.CNN, Along with reliable performance in a variety of fabric textures and lighting conditions, PyTorch is also used to classify fabric and stains. One of the popular CNN architectures used is ResNet-18 which helps deeper networks to overcome the issues of decreasing gradients by utilizing residual connections. Finding the percentage of stain in a fabric is the goal of this project. Image capture and edge extraction should be prioritized, as the textile method for image processing has been studied. According to the tests, this method has a comparatively high detection rate for textile satins.