README.md

MVTec-2K

Google Drive: MVTec-2K.zip
Hugging Face: XimiaoZhang/MVTec-2K

huggingface-cli download --repo-type dataset  XimiaoZhang/MVTec-2K --local-dir MVTec-2K --resume-download

VisA-2K

Google Drive: VisA-2K.zip
Hugging Face: XimiaoZhang/VisA-2K

huggingface-cli download --repo-type dataset  XimiaoZhang/VisA-2K --local-dir VisA-2K --resume-download

MVTec-4K

Google Drive: MVTec-4K.zip
Hugging Face: XimiaoZhang/MVTec-4K

huggingface-cli download --repo-type dataset  XimiaoZhang/MVTec-4K --local-dir MVTec-4K --resume-download

RealIAD-2K

First apply for the Real-IAD dataset.

Download the Real-IAD dataset using the following command:

huggingface-cli login    # Login your account
huggingface-cli download --repo-type dataset Real-IAD/Real-IAD realiad_raw/bottle_cap.zip --local-dir ./ --resume-download
huggingface-cli download --repo-type dataset Real-IAD/Real-IAD realiad_raw/mint.zip --local-dir ./ --resume-download
huggingface-cli download --repo-type dataset Real-IAD/Real-IAD realiad_raw/usb_adaptor.zip --local-dir ./ --resume-download

Unzip them and get the following directory structure:

    |--data                         
        |--RealIAD-2K            
            |--bottle_cap
              |--test.jsonl
              |--train.jsonl
            |--...
            |--test_uni.jsonl   
            |--train_uni.jsonl   
        |--realiad_raw
            |--bottle_cap
              |--NG
              |--OK
            |--...
        |--create_realiad2k.py

Run python create_realiad2k.py to create the dataset.

Please place the downloaded datasets in this folder.

If you are using anomaly synthesis-based methods such as DeSTSeg, RealNet, or RD++, you will also need to download the DTD dataset.

The complete directory structure is as follows:

    |--data                         
        |--MVTec-2K            
            |--bottle
            |--capsule
            |--...
            |--test_uni.jsonl   
            |--train_uni.jsonl   
        |--RealIAD-2K
            |--bottle_cap
            |--mint
            |--...
            |--test_uni.jsonl   
            |--train_uni.jsonl      
        |--...    
        |--DTD          
            |--images
            |--labels
            |--imdb