R3D-AD: Reconstruction via Diffusion for 3D Anomaly Detection

Website | Paper

R3D-AD: Reconstruction via Diffusion for 3D Anomaly Detection,
Zheyuan Zhou∗, Le Wang∗, Naiyu Fang, Zili Wang, Lemiao Qiu, Shuyou Zhang ECCV 2024

Installation

pip install easydict faiss-gpu ninja numpy open3d==0.16.0 opencv-python-headless pyyaml scikit-learn scipy tensorboard timm torch tqdm 
pip install "git+https://github.com/unlimblue/KNN_CUDA.git#egg=knn_cuda&subdirectory=."
pip install "git+https://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib"

Datasets

Anomaly-ShapeNet

Download dataset from google drive and extract pcd folder into ./data/shapenet-ad/

shapenet-ad
├── ashtray0
    ├── train
        ├── ashtray0_template0.pcd
        ...
    ├── test
        ├── ashtray0_bulge0.pcd
        ...
    ├── GT
        ├── ashtray0_bulge0.txt
        ... 
├── bag0
...
...
├── vase9

Training and Testing

python train_test.py PATH_TO_CONFIG

Visualization

python vis_result.py PATH_TO_LOGS

Acknowledgement

Thanks to previous open-sourced repo:

PVD

diffusion-point-cloud

Citation

If you find this project useful in your research, please consider cite:

@inproceedings{zhou2024r3dad,
  title={R3D-AD: Reconstruction via Diffusion for 3D Anomaly Detection},
  author={Zhou, Zheyuan and Wang, Le and Fang, Naiyu and Wang, Zili and Qiu, Lemiao and Zhang, Shuyou},
  booktitle={European Conference on Computer Vision (ECCV)},
  year={2024}
}