: A graph transformer for semantic segmentation of 3D meshes

Giuseppe Vecchio, Luca Prezzavento, Carmelo Pino, Francesco Rundo, Simone Palazzo, Concetto Spampinato

Abstract: Highlights•A two-stream architecture exploiting cluster information combined to Laplacian positional encoding to further enforce locality in mesh segmentation.•Interleaved self-attention and cross-attention operations between mesh triangles and clusters.•State-of-the-art performance on multiple benchmarks for 3D mesh segmentation task.