🌊 Mirage-Test Dataset

Mirage-Test is a modern test-only dataset for benchmarking AI-generated image detection models. It contains real (0_real) and fake (1_fake) images across five distinct content domains, designed to evaluate generalization across diverse visual semantics. The fake images are generated using state-of-the-art generative models specifically optimized for perceptual realism and visual fidelity.

📌 This dataset is for evaluation only. No training split is provided.

📁 Dataset Structure

Images are organized hierarchically by content type and authenticity:

Mirage-Test/
├── Animal/
│ ├── 0_real/ # Real animal photos
│ └── 1_fake/ # AI-generated animal images
├── Anime/
│ └── 1_fake/ # AI-generated anime-style images
├── Human/
│ ├── 0_real/ # Real human photos
│ └── 1_fake/ # AI-generated human images
├── Object/
│ ├── 0_real/ # Real object photos
│ └── 1_fake/ # AI-generated object images
├── Scene/
│ ├── 0_real/ # Real landscape/architecture photos
│ └── 1_fake/ # AI-generated scenes images
├── metadata.parquet
└── README.md

• Total samples: 49000

📥 Downloading Raw Files

To download the dataset with original folder structure:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="Yunncheng/Mirage-Test",
    repo_type="dataset",
    local_dir="./Mirage-Test"
)

🙏 Acknowledgements

• Generated using state-of-the-art diffusion models (e.g., Stable Diffusion, FLUX)
• Real images sourced from publicly available, royalty-free image platforms (e.g., Pexels)

📝 Citation

If you find this work useful for your research, please cite our paper:

@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}
}