SaliencyCut

Code for "SaliencyCut: Augmenting Plausible Anomalies for Anomaly Detection".

Requirements

• matplotlib==3.5.1
• numpy==1.21.5
• pandas==1.3.5
• Pillow==8.4.0
• scikit_learn==1.0.2
• torch==1.9.0
• torchvision==0.10.0
• tqdm==4.64.0

Run

python train.py --dataset_root=./data/mvtec_anomaly_detection --classname='carpet' --nAnomaly=10 --know_class='cut'

• dataset_root denotes the path of the dataset.
• classname denotes the subset name of the dataset.
• nAnomaly denotes the number of anomaly samples involved in training (general setting: 10, hard setting: 1, anomaly-free setting: 0).
• know_class (optional) specifies the anomaly category in the training set to evaluate under hard setting.

Citation

@article{ye2023saliencycut,
  title={SaliencyCut: Augmenting Plausible Anomalies for Open-set Fine-Grained Anomaly Detection},
  author={Ye, Jianan and Hu, Yijie and Yang, Xi and Wang, Qiu-Feng and Huang, Chao and Huang, Kaizhu},
  journal={arXiv preprint arXiv:2306.08366},
  year={2023}
}