---
license: other
library_name: pytorch
pipeline_tag: image-to-3d
tags:
  - 3d
  - gaussian-splatting
  - novel-view-synthesis
  - feed-forward
  - pytorch
---

# OF<sup>3</sup>GS: <u>O</u>n-the-<u>F</u>ly <u>F</u>eed-<u>F</u>orward 3D <u>G</u>aussian <u>S</u>platting from Unposed Images

[Project Website](https://richardchen225.github.io/of3gs/) | [Paper](https://arxiv.org/abs/2606.03254) | [GitHub Repo](https://github.com/richardchen225/OF3GS_code)

This repository hosts the official OF3GS checkpoint.

## Model File

- `model.ckpt`: OF3GS checkpoint for inference and evaluation.

## Usage

Please see the GitHub repository for installation instructions, dataset preparation, and testing commands.

The checkpoint is expected to be used with the FreeStreamGS codebase by setting:

```bash
PRE_FREEGS_PATH=/path/to/model.ckpt
```

or by downloading this file into:

```text
weights/of3gs.ckpt
```

## Citation

If you find OF3GS useful, please cite:

```bibtex
@article{chen2026of3gs,
  title={OF3GS: On-the-Fly Feed-Forward 3D Gaussian Splatting from Unposed Images},
  author={Chen, Ruiyang and Li, Feiran and Zhou, Chu and Li, Zonglin and Ma, Zhanyu and Guo, Heng},
  journal={arXiv preprint arXiv:2606.03254},
  year={2026}
}
```

## License

The code and model are released under the OF3GS license. Please see the license file in the GitHub repository for details.