DFRD: Dual Flow Reverse Distillation for Unsupervised Anomaly Detection

Official PyTorch implementation of DFRD

Datasets

We use the MVTec AD dataset for experiments.

The data directory structure should be:

data
└── mvtec
├── bottle
│ ├── ground_truth
│ ├── test
│ └── train
├── cable
│ ├── ground_truth
│ ├── test
│ └── train
...
└── zipper
├── ground_truth
├── test
└── train

Requirements

• PyTorch 2.0.1
• CUDA 11.8+
• Other dependencies: conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

Testing

1. Run test_visaul.py

Acknowledgement We borrow some codes from RD4AD