ContourCraft is a learning-based solution for handling intersections in neural cloth simulations.
Unlike conventional approaches that critically rely on intersection-free inputs,
ContourCraft robustly recovers from intersections introduced through missed collisions,
self-penetrating bodies, or errors in manually designed multi-layer
outfits.
The technical core of ContourCraft is a novel intersection contour loss that penalizes interpenetrations and
encourages
rapid resolution thereof.
We integrate our intersection loss with a collision-avoiding repulsion objective into a neural cloth simulation
method
based on graph neural networks (GNNs).
We demonstrate our method's ability across a challenging set of diverse multi-layer outfits under dynamic human
motions.
Our extensive analysis indicates that ContourCraft significantly improves collision handling for learned simulation
and
produces visually compelling results.
We extend learned GNN-based garment simulation approach introduced in HOOD with two key modifications. First, we add cloth-cloth correspondences to the graph and use repulsion loss term to prevent intersections. Second, we introduce a novel Intersection Contour loss term that penalizes length of interpenetrations and encourages model to resolve them. Check out our full paper for more details!
Automatic resizing procedures often introduce unwanted intersetions that need to be manually rectified before starting the simulation. ContourCraft can resolve these intersections on the fly and does not require any manual intervention.
Results of our method can be easily exported to industry-standard formats like Alembic for further processing in VFX pipelines.
nt
@inproceedings{grigorev2024contourcraft,
title={{ContourCraft}: Learning to Resolve Intersections in Neural Multi-Garment Simulations},
author={Grigorev, Artur and Becherini, Giorgio and Black, Michael and Hilliges, Otmar and Thomaszewski, Bernhard},
booktitle={ACM SIGGRAPH 2024 Conference Papers},
pages={1--10},
year={2024}
}