Semantic Segmentation of Textured Non-manifold 3D Meshes using Transformers

Mohammadreza Heidarianbaei, Max Mehltretter, Franz Rottensteiner · 2026

Read Semantic Segmentation of Textured Non-manifold 3D Meshes using Transformers by Mohammadreza Heidarianbaei, Max Mehltretter, Franz Rottensteiner (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-01836.txt.

Description

Heidarianbaei and colleagues present a texture transformer for segmenting textured non-manifold 3D meshes, whose irregular structure challenges learning. Recent mesh methods ignore rich textural information. Their model learns from raw pixels per mesh face, with layered multi-scale aggregation. A texture branch turns face-level pixels into a learned token, fused with shape descriptors and processed by Two-Stage Blocks for local and global flow. On the Semantic Urban Meshes benchmark and a new roof-tile dataset annotated for damage, the method reaches 81.9 percent mF1, beating existing approaches.

Catalog record

CREATORMohammadreza Heidarianbaei, Max Mehltretter, Franz Rottensteiner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49257
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.01836

Browse the full library · About the Tower of Babel