Matheus Gadelha
Senior Research Scientist
Seattle, WA
Matheus is a researcher interested in all things related to 3D. His work is focused on models and representations of tridimensional data for both discriminative and generative models with a particular interest in learning directly from images. Lately, he has been focused on using learnings from 3D representation learning to enhance controllability of generative image and video models.
A full list of his publications and more information about his research can be found at his personal website.
Publications
News
A Single Primitive to Rule Them All: SuperFit at CVPR 2026 Compresses 3D Shapes into Editable Building Blocks
SuperFit - presented at CVPR 2026 - converts complex 3D shapes into compact assemblies of simple, interpretable building blocks, called SuperFrusta.
June 2, 2026SIGMA-GEN: Placing any subject, anywhere, exactly where you want it
SIGMA-GEN is an experimental unified framework from Adobe Research, presented at ICLR 2026, for multi-identity-preserving image generation — placing multiple subjects into a scene in a single pass, with each subject's appearance faithfully preserved.
April 1, 2026