Skip to content

VastGaussian

An unofficial large-scene 3D Gaussian reconstruction project

Visit official site
Category AI 3D tools

What VastGaussian is

The project is described as an open-source 3D scene reconstruction effort that uses 3D Gaussians to model the geometry and appearance of large scenes, written from scratch by the author.

What you can do with it

  • Camera-position-based region division
  • Position-based data selection and visibility-based camera selection
  • Coverage-based point selection with decoupled appearance modeling

Application scenarios

  • Applied to the UrbanScene3D dataset
  • Applied to the Mill-19 dataset
  • Applied to the tandt_db dataset

Pros & cons

✓ What we like

  • Handles large datasets and simplifies the original 3DGS project
  • Seamless merging is implemented along with the other modules

! What to watch out for

  • The author notes the from-scratch implementation may contain errors
  • Parameters in the arguments file need adjusting for a dataset

FAQ

Is it an official release?

The material calls it an unofficial implementation of Vast 3D Gaussians for Large Scene Reconstruction.

How is training started?

The material names a training script after setting up the environment.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review