Light and materials estimation in 3D Gaussian Splatting models

Thesis Proposal Details

Supervisor: Simone Milani

Co-supervisor: Chiara Schiavo; Leonardo Monchieri; Elena Camuffo

Co-supervisor Department/Company: Simone Milani

Creation Date: 15/07/2025 22:10

Description

3D Gaussian splatting is a novel technique for real-time rendering of complex 3D scenes. It represents a scene as a collection of 3D Gaussian primitives, which are efficiently rendered to create high-quality, view-dependent images with impressive detail and speed.

The candidates that are interested on this thesis topics will focus on:

1. Estimate the lighting on the multi-view images used in Gaussian Splatting reconstructions.
2. Estimate the material properties and textures in Splatting reconstructions.

Thesis require a preliminary knowledge of Computer Vision and Python programming.

Dataset and methods

Dataset type: Already acquired data

Dataset description: dataset of pictures taken from different viewpoints in a given static scene

List of Methods: 3D Gaussian splatting, semantic segmentation, point cloud reconstruction

Preparatory Courses

3D Vision & eXtended Reality

Tags
3D Gaussian Reconstruction Splatting immersive rendering segmentation technologies
Back to proposals list