This repository stores the preprocessed 832-size assets used by the public
ForgeryVCR training and evaluation workflow.
datasets/
├── train/
│ ├── sft/ # CASIA v2 archives for expert selection and Agent SFT
│ └── rl/ # IMD2020 and FantasticReality archives for GRPO
└── test/ # public evaluation benchmark archives
Tampered-set JSON files use paths relative to their own dataset directory, for
example ./forged/example.png and ./gt/example_gt.png.… See the full description on the dataset page: https://huggingface.co/datasets/ForgeryVCR-Team/ForgeryVCR.
