---
license: mit
tags:
- ai-generated images
- ai-generated image detection
- test-set
- deepfake
- forgery-detection
- computer-vision
task_categories:
- image-classification
language:
- en
dataset_info:
  features:
    - name: file_name
      dtype: string
      description: "Relative path to the image under root."
    - name: image
      dtype: image
    - name: is_real
      dtype: string
    - name: content_type
      dtype: string
  data_files:
    - split: test
      path: test.parquet
---

# 🌊 Mirage-Test Dataset

[![Paper](https://img.shields.io/badge/arXiv-2511.08423-B31B1B.svg)](https://arxiv.org/abs/2511.08423)
[![GitHub](https://img.shields.io/badge/GitHub-Code-181717.svg?logo=github)](https://github.com/yunncheng/OmniAID)
[![Hugging Face Models](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-yellow)](https://huggingface.co/Yunncheng/OmniAID/tree/main)
[![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/Yunncheng/OmniAID-Demo)
[![License](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)

**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:

```bash
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:

```python
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](https://github.com/Stability-AI/stablediffusion), [FLUX](https://github.com/black-forest-labs/flux))
- Real images sourced from publicly available, royalty-free image platforms (e.g., [Pexels](https://www.pexels.com/))


## 📝 Citation

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

```bibtex
@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}
}
```