fudan-zvg/4d-gaussian-splatting

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1025 stars · 82 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official PyTorch/CUDA implementation of 4D Gaussian Splatting (ICLR 2024), which represents and renders dynamic scenes in real time using 4D Gaussian primitives with a dedicated rendering pipeline. Built on top of 3D Gaussian Splatting, it supports training on datasets like DyNeRF and D-NeRF for photorealistic dynamic novel view synthesis.

Use cases

• render dynamic scenes in real time from multi-view video
• train a 4D gaussian splatting model on the DyNeRF dataset
• do dynamic novel view synthesis from monocular video
• reconstruct dynamic 3D scenes with gaussian primitives
• reproduce the ICLR 2024 4DGS paper results
• create bullet-time or free-viewpoint video effects

When to choose

• you need real-time photorealistic rendering of dynamic scenes
• you want the official, citable implementation of the 4D Gaussian Splatting paper
• you are doing research on dynamic scene representation or novel view synthesis
• you already have a CUDA-capable GPU and are comfortable with the 3D Gaussian Splatting toolchain

When to avoid

• you only need static scene reconstruction (plain 3D Gaussian Splatting is simpler)
• you have no CUDA GPU, since training and rendering require one
• you need a production-ready end-user application rather than research code
• you need cross-platform or CPU-only rendering

Facets

library · maturity active

graphics machine-learning image-processing simulation computer-vision graphics deep-learning machine-learning python cpp gaussian-splatting novel-view-synthesis neural-rendering 4d-reconstruction research-code iclr-2024 linux gpu cuda

Member repositories

For agents

markdown · JSON · MCP: product_card(name="fudan-zvg/4d-gaussian-splatting")