COSMI: Compositional Synthesis of Multi-Object Interactions
COSMI composes single-object captures into multi-object interactions and trains a single diffusion model that generates up to five objects and generalizes to unseen combinations.
I am a Master’s student in Machine Learning at the University of Tübingen, in the Real Virtual Humans group led by Prof. Dr. Gerard Pons-Moll. My research aims to build virtual humans that look and act like real people, with a focus on human–object interaction and layered clothing. I am supported by the Zuse School ELIZA Research-Oriented Master’s Scholarship.
I am seeking a PhD in Virtual Humans, Embodied AI, and 3D Computer Vision.
COSMI composes single-object captures into multi-object interactions and trains a single diffusion model that generates up to five objects and generalizes to unseen combinations.
DAMA reconstructs physically plausible layered avatars of clothed humans from multi-view images, with separated garment layers that can be stacked, reordered, and simulated.
Lightweight visualizer for Gaussian head avatars compatible with PhysHead (CVPR 2026), enabling interactive editing with FLAME-driven control, hair dynamics, coloring, cutting, and hairstyle transfer.