README.md

Continual Anomaly Detection

Official code for ACMMM 2022 paper:

Title: Towards Continual Adaptation in Industrial Anomaly Detection [pdf].

Datasets

To train on the MVTec Anomaly Detection dataset download the data and extract it. For the additional Magnetic Tile Defects dataset, we download the data then run datasets/utils/make_mtd_ano.py for anomaly detection.

Enviroment setup

pip install -r requirements.txt

Getting pretrained ViT model

ViT-B/16 model used in this paper can be downloaded at here.

Run

We provide the configuration file to run CAD on multiple benchmarks in configs.

python main.py --config-file ./configs/cad.yaml  --data_dir ../datasets/mvtec --mtd_dir ../datasets/mtd_ano_mask

You can run the method you need by modifying the configuration file.