README.md

metadata

title: PoseShield Collision Fix
emoji: 🛡️
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
short_description: Fix self-collisions in SMPL-H poses via neural fields
python_version: '3.12'
startup_duration_timeout: 30m

PoseShield: Neural Collision Fields for Human Self-Collision Resolution

This Space demonstrates PoseShield, a post-hoc self-collision resolver for SMPL-H poses. Given a self-intersecting pose, it uses a learned neural collision field as a differentiable constraint and solves a constrained optimization (SLSQP) to find a nearby collision-free pose.

Usage

1. Upload a .pkl file containing a SMPL-H pose with a pose key (21 joints × 3 axis-angle rotations, shape (63,)), or try one of the provided example poses.
2. Click "Resolve Collisions" to run the optimization.
3. View the before/after 3D mesh visualizations and optimization metrics.

References

• Paper (arXiv:2606.29686)
• GitHub
• Model Weights