jayin92/Skyfall-GS
Skyfall-GS: Synthesizing Immersive 3D Urban Scenes from Satellite Imagery
Combines satellite imagery with open-domain diffusion models to generate large-scale 3D urban scenes without 3D annotations, using a two-stage pipeline: initial reconstruction via 3D Gaussian splatting, then iterative dataset update (IDU) with curriculum-driven refinement to enhance geometry and textures. Supports real-time rendering and exploration with cross-view consistency, integrating COLMAP reconstructions and custom satellite data preprocessing via the SatelliteSfM pipeline.
762 stars. Actively maintained with 3 commits in the last 30 days.
Stars
762
Forks
77
Language
Python
License
Apache-2.0
Category
Last pushed
Mar 18, 2026
Commits (30d)
3
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/jayin92/Skyfall-GS"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related models
ActiveVisionLab/gaussctrl
[ECCV 2024] GaussCtrl: Multi-View Consistent Text-Driven 3D Gaussian Splatting Editing
Tencent-Hunyuan/Hunyuan3D-2
High-Resolution 3D Assets Generation with Large Scale Hunyuan3D Diffusion Models.
deepseek-ai/DreamCraft3D
[ICLR 2024] Official implementation of DreamCraft3D: Hierarchical 3D Generation with...
caiyuanhao1998/Open-DiffusionGS
Baking Gaussian Splatting into Diffusion Denoiser for Fast and Scalable Single-stage Image-to-3D...
ai4ce/GARF
[ICCV2025] GARF: Learning Generalizable 3D Reassembly for Real-World Fractures