AdamYao/3D_Visual_Illusion_Depth_Estimation_Real

3D Visual Illusion Depth Estimation Dataset Real

Dataset Summary

The 3D Visual Illusion Depth Estimation Dataset Real is designed for research on stereo and monocular depth estimation in 3D visual illusion scenes. It contains left and right stereo images, disparity map, and illusion-region masks.

Dataset Structure

Each sample in the dataset includes:

• —left: Left-view RGB image
• —right: Right-view RGB image
• —depth: Monocularly estimated depth map
• —mask: Binary mask indicating illusion regions

├── disp
│    └── scene_name
│           └── xxx.pfm
├── left
│    └── scene_name
│           └── xxx.png
├── mask
│    └── scene_name
│           └── xxx-illusion.jpg
└── right
     └── scene_name
            └── xxx.png

Data Statistics

The current release of the dataset contains:

• —Total size: [ 8 ] GB
• —Number of videos: [ 83 ]
• —Number of image pairs: [ 455 ]

Repository

The dataset and related code are available on GitHub: 👉 3D-Visual-Illusion-Depth-Estimation

Citation

If you find this project useful, please consider citing:

@article{yao20253dvisualillusiondepth,
  title={3D Visual Illusion Depth Estimation},
  author={Chengtang Yao and Zhidan Liu and Jiaxi Zeng and Lidong Yu and Yuwei Wu and Yunde Jia},
  journal={arXiv:2505.13061},
  year={2025}
}