hustvl/4DGaussians

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

3895 stars · 385 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

An official PyTorch implementation of 4D Gaussian Splatting (4D-GS) for real-time rendering of dynamic scenes, published at CVPR 2024. It combines 3D Gaussians with 4D neural voxel encoding to train high-resolution dynamic scenes in ~30 minutes and render at 80+ FPS on a single RTX 3090.

Use cases

• render dynamic 3D scenes in real time
• novel view synthesis of moving scenes from video
• train a 4D gaussian splatting model on D-NeRF or HyperNeRF datasets
• reproduce CVPR 2024 4D-GS research results
• compare dynamic scene rendering methods against radiance fields

When to choose

• you need real-time rendering of dynamic scenes with high quality
• you want a fast-training alternative to per-frame 3D Gaussian Splatting
• you are doing research on neural rendering or radiance fields

When to avoid

• you only need static scene reconstruction (plain 3D Gaussian Splatting suffices)
• you lack a CUDA GPU or cannot build custom rasterization submodules
• you need a production-ready application rather than research code

Facets

library · maturity active

graphics computer-vision machine-learning simulation computer-vision graphics deep-learning python cli gaussian-splatting novel-view-synthesis dynamic-scenes neural-rendering radiance-field cvpr-2024 research-code gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="hustvl/4DGaussians")