Graph-based neural simulators like HOOD (CVPR 2023) are excellent at handling collisions between the cloth and the human body. However, they often struggle with Self-Collisions—instances where the ...

We study whether a neural model can replace an explicit physics simulator by learning dynamics directly from visual observations. The framework couples a VAE for image–latent translation with a causal ...

Abstract: Comparing robotic cloth-manipulation systems in a real-world setup is challenging. The fidelity gap between simulation-trained cloth neural controllers and real-world operation hinders the ...