Gaussian Splatting 3 Ways
4 months ago
- #COLMAP-PyTorch
- #Tkinter-OpenGL
- #3D-Gaussian-Splats
- Tkinter + OpenGL desktop app for training and viewing 3D Gaussian splats from casual captures.
- Uses COLMAP for camera poses, PyTorch + gsplat for training, and stores everything in a reproducible cache tree.
- Supports three methods: Traditional COLMAP + gsplat, Depth Anything 3 3D Gaussian Estimation, and SHARP Monocular View Synthesis.
- App organized around a core state object (AppState), four UI tabs (Inputs, COLMAP, Training, Exports), and a backend pipeline.
- Cache layout includes per-scene inputs/outputs under cache for repeatable workflows.
- Training times vary: gsplat (~5 min), Depth Anything 3 (~3-4 min), SHARP (~2.5 min).
- Includes a portable bundle builder (build.bat) for self-contained deployment.
- Requirements: Windows/Linux, Python 3.10+, GPU with CUDA, ffmpeg, COLMAP binaries.
- Known issues: Thumbnails loading, splat rendering, camera control, and CUDA SDK path requirements.