| Wenjia Wang | 王文佳 |
I am a researcher working on human-centric models and humanoid intelligence. I obtained my Ph.D. in Computer Science from the University of Hong Kong in November 2026, under the supervision of Prof. Taku Komura.
At the Miro-CGVU Joint Lab, I lead research and recruit and mentor research assistants for collaborative projects in embodied intelligence, offering opportunities to prepare for future MPhil or Ph.D. study. Please contact me if you are interested.
Research Interests
My research connects human prior models, world models, and humanoid intelligence.
Human Prior Model
Learning human representations from perception, motion and physical interactions.
World Model
Modeling 3D worlds and characters through generative dynamics and animation.
Humanoid Intelligence
Learning human-scene interaction, task planning and physically grounded whole-body behavior.
Recent News
- Nov 5, 2026 [Ph.D. Defense] Successfully defended my Ph.D. thesis at HKU. You can now call me Dr. Wang!
- Apr 8, 2026 [Talk] Gave a TechBeat talk on EmbodMocap.
- Mar 10, 2026 [Open Source] Open-sourced EmbodMocap.
- Feb 22, 2026 [Accept] EmbodMocap accepted to CVPR 2026.
- Jan 6, 2026 [Project Page] Project page online for EmbodMocap.
- Jan 6, 2026 [Preprint] EgoGrasp preprinted on arXiv.
Selected Papers
EmbodMocap ConfCVPR 2026
Reconstructs human motion and scenes using two moving iPhones, preserving the spatial relationships needed for scene-aware motion tracking and robot learning.
SIMS ConfICCV 2025
Connects high-level planning with physics-based control through retrieval-augmented LLM scripts for long-horizon, stylized human-scene interactions.
Zolly ConfICCV 2023, Oral 🏆️
Introduces distortion-aware representations and camera modeling for accurate human mesh recovery from perspective-distorted images.
Academic Services
- Teaching Assistant: COMP7404C (Spring 2024, HKU); COMP3329 (Spring 2024, HKU)
- Conference Reviewer: ICCV (2023); ICLR (2024–2026); ACCV (2024); SIGGRAPH Asia (2024–2026); CVPR (2025–2026); 3DV (2025–2026); ECCV (2026); SIGGRAPH (2026); BMVC (2026); NeurIPS E&D Track (2026).
- Journal Reviewer: IEEE TIP (2022); IEEE TVCG (2026).
Invited Talks
- Apr 2026 低成本易扩展的动态双目场景与动作重建方案, TechBeat 人工智能社区.