caiyuanhao1998/Open-DiffusionGS

Baking Gaussian Splatting into Diffusion Denoiser for Fast and Scalable Single-stage Image-to-3D Generation and Reconstruction (ICCV 2025)

49
/ 100
Emerging

Integrates Gaussian Splatting directly into the diffusion denoising process to enable single-image 3D reconstruction in ~6 seconds without requiring separate multi-view diffusion or depth estimators. Supports both object-level generation from single views with mesh export and scene-level reconstruction, trained on Objaverse and RealEstate10K datasets using diffusion-gaussian-rasterization and simple-knn backends. Achieves 7.5× speedup over comparable methods while maintaining quality across diverse inputs from synthetic objects to in-the-wild photographs.

823 stars.

No Package No Dependents
Maintenance 10 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 13 / 25

How are scores calculated?

Stars

823

Forks

38

Language

Python

License

MIT

Last pushed

Jan 28, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/caiyuanhao1998/Open-DiffusionGS"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.