PhysSkin++: Real-Time and Generalizable Physics-Based Animation via Self-Supervised Neural Skinning

One-line takeaway

ZJU's PhysSkin++ learns continuous neural skinning fields from static geometry to build compact deformation subspaces for real-time physics-based animation that generalizes across shapes and discretizations — no per-object training and no simulation trajectories, via physics-informed self-supervised learning.

浙大 3D 视觉组的 PhysSkin++ 从静态几何学习连续神经蒙皮场，构造紧凑的形变子空间，实现跨形状、跨离散方式的实时物理动画；不需逐物体训练，也不需仿真轨迹，用物理约束的自监督学习训练。