Thrilled to share our #SIGGRAPHAsia2025 work, PhysHMR: Learning Humanoid Control Policies from Vision for Physically Plausible Human Motion Reconstruction (https://t.co/1LUl9oBC28).
Thrilled to share our #SIGGRAPHAsia2025 work, PhysHMR: Learning Humanoid Control Policies from Vision for Physically Plausible Human Motion Reconstruction (https://t.co/1LUl9oBC28).
✨ Highlights:
• End-to-end visual→action (no multi-stage pipelines, less error accumulation)
• Efficient training via supervised distillation + RL (PPO)
• Real-world transfer: the policy can also drive a Unitree G1 robot 🤖
Check the videos 👇