Semantic Gaussian Splatting

Live Demo Link: https://shahan.ca/sgs

This project is a fork of gaussian splatting that adds semantic instance information to every single gaussian which is optimized during normal training by doing some extra training.

You can view a demo and mode detailed explaination of how it works here:

Videos

Photos