This repository contains the complete code for a competition solution on cloth defect detection using computer vision. The approach employs Cascade-RCNN, a multi-stage object detection framework based on Faster R-CNN, combined with a change detection methodology that merges test images with template images to identify fabric defects. The solution was developed in Python and ranked 3rd in the 2019 Guangdong Industrial Intelligent Manufacturing Innovation Competition hosted on Alibaba Cloud’s Tianchi platform.