FALCO-WAFER : Feature-Aware Lightweight Contextual Detector for Wafer Defects Detection

PyTorch implementation for ITC-Asia：FALCO-WAFER : Feature-Aware Lightweight Contextual Detector for Wafer Defects Detection

Intuitive illustration of class-generalizable anomaly detection and conceptual illustration of residual feature learning:

---

Note: This repository contains the primary implementation of our ITCA paper, developed based on the Ultralytics framework. While it may not be a line-by-line replica, the core logic and components faithfully represent the method described in the paper.

Installation

This project is developed based on the Ultralytics framework (version 8.0.201). To configure the base environment, please refer to the official Ultralytics documentation:👉 https://docs.ultralytics.com

This project is tested and recommended to run with the following core dependencies:

• torch 2.0.0+cu117
• torchvision 0.15.0+cu117
• mmengine 0.10.7
• mmcv 2.1.0

Download Datasets

Please download dataset from Texture-AD. To make it compatible with YOLO-based detectors, we provide a format conversion script MVTEC2YOLO.py for preprocessing.

Training and Evaluating

• Run code for training, you can use you own config in train.py

python train.py 

• Run code for evaluating, you can use you own config in val.py

python val.py 

• Run code for testing, you can use you own config in test.py

python test.py