Yan Zhang

Yan Zhang is a scientific leader at Meshcapade and a guest lecturer at ETH Zurich's Computer Vision and Learning Group (VLG). He previously served as a postdoctoral researcher at ETH Zurich (2020-2023) and research intern at Max Planck Institute for Intelligent Systems (2018-2020).

His research focuses on generative human foundation models, human motion and behavior synthesis, 3D human perception, and applications in AR/VR, embodied AI, and interactive avatars. He has pioneered methods for scene-conditioned motion generation, contact-aware reconstruction, and egocentric interaction modeling.

His recent publications (2025-2020) span

• Real-time motor models for avatars (PRIMAL, ICCV'25)
• Diffusion architectures for motion (RoHM, CVPR'24)
• Scene-population algorithms (Odysseus, CVPR'22)
• Physics-aware reconstruction (EgoHMR, ICCV'23)
• Whole-body grasping models (SAGA, ECCV'22)
• Multi-modal datasets (EgoBody, ECCV'22)

Scientific recognition includes the Qualcomm Innovative Fellowship Europe 2023. He organized workshops at CVPR'25, ECCV'24, and ECCV'22, and served on senior program committees (AAAI'26) and area chairs (CVPR'25).

As co-supervisor, he mentored student projects on diffusion-based hand motion capture, 3D pose estimation, body-scene interaction, and mixed reality navigation at ETH Zurich (2020-2023). His work bridges computer vision, machine learning, and computer graphics to advance human-centric AI systems.