G³Splat: Geometrically Consistent Generalizable Gaussian Splatting

G³Splat is a pose-free, self-supervised framework for generalizable Gaussian splatting, achieving strong performance in:

• 🎯 Geometry Reconstruction (depth / mesh)
• 📐 Relative Pose Estimation (no poses required at inference)
• 🎨 Novel View Synthesis

Start here

• Models / checkpoints: g3splat/g3splat
• Code + demos: https://github.com/m80hz/g3splat
• Project page: https://m80hz.github.io/g3splat
• Paper: https://arxiv.org/abs/2512.17547

Citation

@inproceedings{g3splat,
  title     = {G3Splat: Geometrically Consistent Generalizable Gaussian Splatting},
  author    = {Hosseinzadeh, Mehdi and Chng, Shin-Fang and Xu, Yi and Lucey, Simon and Reid, Ian and Garg, Ravi},
  booktitle = {arXiv:2512.17547},
  year      = {2025},
  url       = {https://arxiv.org/abs/2512.17547}
}

Contact / Maintainers

• Maintained by the G³Splat authors.
• For questions and issues, please use the GitHub issue tracker: https://github.com/m80hz/g3splat/issues