OmniAID

OmniAID Checkpoints

This repository hosts the released checkpoints and corresponding training configs for OmniAID.

Companion repositories:

• Yunncheng/OmniAID-Demo — interactive Gradio Space that consumes these weights.
• Yunncheng/OmniAID — full training code (the configs here mirror config/ in the training repo).

Repository Layout

OmniAID-ckpt/
├── ckpt/                              # Model weights (HuggingFace LFS)
│   ├── checkpoint_omniaid_dino_v1.pth
│   ├── checkpoint_omniaid_dino_v2.pth
│   ├── checkpoint_omniaid_v1.pth
│   ├── checkpoint_omniaid_v2.pth
│   └── checkpoint_omniaid_genimage_sd14.pth
└── config/                            # Training / inference configs
    ├── config_omniaid_dino_v1.json
    ├── config_omniaid_dino_v2.json
    ├── config_omniaid_v1.json
    ├── config_omniaid_v2.json
    └── config_omniaid_genimage_paper.json

Checkpoints

The _v2 suffix marks the current public release.

Downloading Weights

from huggingface_hub import hf_hub_download

ckpt_path = hf_hub_download(
    repo_id="Yunncheng/OmniAID",
    filename="ckpt/checkpoint_omniaid_dino_v2.pth",
)
state_dict = torch.load(ckpt_path, map_location="cpu", weights_only=False)
if "model" in state_dict:
    state_dict = state_dict["model"]

Pair the checkpoint with its config:

Citation

@article{guo2025omniaid,
  title={OmniAID: Decoupling Semantic and Artifacts for Universal AI-Generated Image Detection in the Wild},
  author={Guo, Yuncheng and Ye, Junyan and Zhang, Chenjue and Kang, Hengrui and Fu, Haohuan and He, Conghui and Li, Weijia},
  journal={arXiv preprint arXiv:2511.08423},
  year={2025}
}