Towards Total Recall in Industrial Anomaly Detection

Here just a unofficial implement of PatchCore.

The original paper can be find on: https://arxiv.org/abs/2106.08265

requirements

opencv pytorch faiss scikit-learn tqdm matplotlib

把正常图片放一个文件夹,然后运行 train.py 生成memory bank 到compare文件夹,比如:

python train.py --img_dir yourImageDir --save_dir memoryBankSaveDir

然后运行 test.py 加载上一步的memory_bank.npy 测试测试数据

python test.py --img_dir testImageDir --memory_bank memoryBankSaveDir