DM-GRD

💡 This is the official implementation of the paper "Dual Memory Networks Guided Reverse Distillation for Unsupervised Anomaly Detection (ACCV 2024)"

🔧 Installation

Using conda

Using environment.yml

conda env create -f envs/environment.yml
conda activate anomaly
pip install -e .

Using requirements.txt

conda create --name anomaly python=3.10.12
conda activate anomaly
pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117
pip install -r envs/requirements.txt
pip install -e .

Using uv

uv venv anomaly --python=3.10.12
source anomaly/bin/activate
uv pip install -e .

🏆 Dataset preparation

For the MVTec dataset, please download it from this link.

For the BTAD dataset, please download it from this repository.

For the VisA dataset, please download it from this repository.

For the DTD dataset, please download it from this link.

🚀 Experiments

🌞 Training and testing DM-GRD on the MVTec dataset

bash scripts/train_mvtec.sh
bash scripts/test_mvtec.sh

🌞 Training and testing DM-GRD on the BTAD dataset

bash scripts/train_btad.sh
bash scripts/test_btad.sh

🌞 Training and testing DM-GRD on the VisA dataset

bash scripts/train_visa.sh
bash scripts/test_visa.sh

🔗 Citation

If you find our work useful, please cite the following:

@inproceedings{tran2024dual,
  title={Dual Memory Networks Guided Reverse Distillation for Unsupervised Anomaly Detection},
  author={Tran, Chi Dai and Pham, Long Hoang and Tran, Duong Nguyen-Ngoc and Ho, Quoc Pham-Nam and Jeon, Jae Wook},
  booktitle={Proceedings of the Asian Conference on Computer Vision},
  pages={2650--2666},
  year={2024}
}

☎️ Contact

If you have any questions, feel free to contact Chi D. Tran (ctran743@gmail.com or tdc2000@skku.edu).

🙏 Acknowledgement

Our framework is built using multiple open source, thanks for their great contributions.

• hq-deng/RD4AD
• tientrandinh/Revisiting-Reverse-Distillation