Procedural Dataset Generation for Zero-Shot Stereo Matching

We introduce Infinigen-Stereo, a procedural dataset generator specifically optimized for zero-shot stereo matching performance. Using our generator, we create and release Infinigen-Stereo-150k, a new training dataset for stereo matching.

If you find Infinigen-Stereo useful for your work, please consider citing our academic paper:

Procedural Dataset Generation for Zero-Shot Stereo Matching

David Yan, Alexander Raistrick, Jia Deng

@misc{yan2025proceduraldatasetgenerationzeroshot,
      title={Procedural Dataset Generation for Zero-Shot Stereo Matching}, 
      author={David Yan and Alexander Raistrick and Jia Deng},
      year={2025},
      eprint={2504.16930},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2504.16930}, 
}

Generator Code

Coming Soon!

InfinigenStereo Dataset

Our dataset is now available on HuggingFace. You can download it with the command

pip install huggingface-cli
huggingface-cli download pvl-lab/InfinigenStereo --repo-type dataset

The dataset file structure is as follows:

.
└── InfinigenStereo/
    ├── indoor/
    │   └── seed_num/
    │       └── frames/
    │           ├── Image/
    │           │   ├── camera_0
    │           │   └── camera_1
    │           ├── camview/
    │           │   ├── camera_0
    │           │   └── camera_1
    │           ├── disparity/
    │           │   └── camera_0
    │           ├── occ_mask/
    │           │   └── camera_0
    │           └── sky_mask/
    │               └── camera_0
    ├── flying/
    │   └── ...
    └── nature/
        └── ...

Camera 0 and 1 correspond to left and right camera frames, respectively. We provide disparity, occlusion, sky-region masks for the left camera. camview contains .npz files that contain a dictionary with indices K, T, HW, corresponding to calibration, translation, and resolution matrices.