F. Rottensteiner

By explicitly incorporating textural information into a transformer-based architecture, this approach achieves state-of-the-art semantic segmentation on textured 3D meshes, unlocking more accurate scene understanding from complex geometric data.