Texturing 3D Models with Words

Paper Title:

MatAtlas: Text-driven Consistent Geometry Texturing and Material Assignment

Published on:

3 April 2024

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Duygu Ceylan,

Valentin Deschaintre,

Thibault Groueix,

Rosalie Martin,

Chun-Hao Huang,

Romain Rouffet,

Vladimir Kim,

Gaëtan Lassagne

Bullets

Key Details

•

Leverages large scale text-to-image models to texture 3D assets

•

Uses grid pattern diffusion for consistent texturing across views

•

Refines textures over multiple passes to improve quality

•

Assigns retrieved parametric materials for relighting and editing

•

Significantly outperforms prior state-of-the-art texturing methods

Explore the topics in this paper

3d texturing

ai materiai textures

neural texturing

text to texture generation

text-prompted diffusion models

AI generated summary

Texturing 3D Models with Words

This paper presents a method to generate high quality, editable textures on 3D models using text prompts and large language models. It textures models by generating images from multiple views using a grid pattern diffusion strategy for consistency. The textures are refined over multiple passes and used to retrieve parametric materials, enabling relighting and editing.