This is the implementation of paper “Learning Semantic Context from Normal Samplesfor Unsupervised Anomaly Detection”.
Dataset should be stored in ./dataset/xxx.
The Mvtec AD dataset could be download at : https://www.mvtec.com/company/research/datasets/mvtec-ad/
Python = 3.6.8
Pytorch = 1.1.0
opencv = 3.4.2
mmcv = 0.4.0
scikit-image = 0.15.0
On Mvtec AD dataset:
Download checkpoints form https://pan.baidu.com/s/1TEwc_C9HPs6GOrRA1z_HfA code:9s47.
Then unzip file and move it to ./ckpt.
Run this comand to get result of ${class_name} such as bottle etc.
python test.py configs/config_mvtec.py --subset ${class_name}