# CleanFD Dataset

Fake image detection dataset for training and evaluation.

## Download & Extract

```bash
# huggingface_hub 설치 (필요시)
pip install huggingface_hub

# 다운로드
huggingface-cli download leekwoon/cleanfd-dataset --repo-type dataset --local-dir ./cleanfd_data

# 무결성 확인 (선택사항)
cd cleanfd_data
md5sum -c checksums.md5

# 파일 합치기 및 압축 해제
cat data.tar.gz.part_* | tar -xzvf -
```

## Directory Structure

```
data/
├── train/
│   ├── real/
│   │   ├── coco/
│   │   └── lsun/
│   └── fake/
│       ├── aligned/
│       ├── ldm/
│       ├── sd21_inpainted_*/
│       └── ...
├── val/
│   ├── real/
│   └── fake/
├── test/
│   ├── real/redcaps/
│   └── fake/ (sd, Midjourney, flux, etc.)
└── test_processed/
    ├── real/
    └── fake/
```

## Train Data Sources

- **real**: COCO, LSUN
- **fake**:
  - aligned (Rajan aligned inversions)
  - ldm (Latent Diffusion Model)
  - sd21_inpainted_* (B-Free inpainting data)
  - lsun_inpaint_* (LSUN inpainting)

## Test Data Sources

- **real**: RedCaps
- **fake**: SD, Midjourney, Kandinsky, Playground, PixelArt, LCM, Flux, Wuerstchen, Amused, Chameleon, Loki, WildRF