← ArchivePaper2022

N-Cloth: Predicting 3D Cloth Deformation with Mesh-Based Networks

Yudi Li, Min Tang, Yun Yang, Zi Huang, Ruofeng Tong, Shuangcai Yang, Yao Li, Dinesh Manocha

Graph convolution on arbitrary-topology cloth and obstacle meshes predicts plausible 3D cloth deformation at 30-45 fps for up to 100K triangles.

Open the paper ↗Trace in the lineage graph

Abstract

We present a novel mesh‐based learning approach (N‐Cloth) for plausible 3D cloth deformation prediction. Our approach is general and can handle cloth or obstacles represented by triangle meshes with arbitrary topologies. We use graph convolution to transform the cloth and object meshes into a latent space to reduce the non‐linearity in the mesh space. Our network can predict the target 3D cloth mesh deformation based on the initial state of the cloth mesh template and the target obstacle mesh. Our approach can handle complex cloth meshes with up to 100K triangles and scenes with various objects corresponding to SMPL humans, non‐SMPL humans or rigid bodies. In practice, our approach can be used to generate plausible cloth simulation at 30, 45 fps on an NVIDIA GeForce RTX 3090 GPU. We highlight its benefits over prior learning‐based methods and physically‐based cloth simulators.

How to read this

Category: Method: mesh-based neural cloth deformation predictor
Contributions: Graph-convolution network that maps arbitrary-topology cloth and obstacle meshes into a latent space to reduce mesh-space non-linearity
Predicts target 3D cloth deformation from the cloth template's initial state and the target obstacle mesh
Handles complex cloth (up to 100K triangles) and SMPL/non-SMPL humans and rigid bodies at interactive rates (30 to 45 fps on an RTX 3090)
Context: Sits among learning-based cloth methods such as Bertiche et al. Neural Cloth Simulation (2022), differentiating by handling arbitrary mesh topologies via graph convolution.Builds on: Neural Cloth Simulation
Correctness: Claims plausible deformation and benefits over prior learning-based methods and physically-based simulators on the tested cloth and obstacle meshes; results are plausible predictions, not guaranteed physically accurate, and the reported fps is hardware-specific.
Clarity: Moderately technical; a first pass gives the graph-convolution latent idea, a second pass covers the network and topology handling.
How to read it: First pass for the arbitrary-topology graph-convolution framing; second pass on the latent-space mapping and the speed/accuracy tradeoff versus PBS if performance matters to you.

Builds on

Neural Cloth Simulation 2022

Built upon by

Nothing yet.

Related work

GarMatNet: A Learning-Based Method for Predicting 3D Garment Mesh with Parameterized Materials 2021 / MIG
SMPLicit: Topology-aware Generative Model for Clothed People 2021 / CVPR
A Pixel-Based Framework for Data-Driven Clothing 2020 / SCA
3D Hair Synthesis Using Volumetric Variational Autoencoders 2018 / SIGGRAPH Asia

Keywords

n-cloth cloth deformation prediction graph convolution gcn encoder-decoder network fusion network latent space top-k-pooling triangle mesh smpl tailornet penetration loss self-penetration multilayer perceptron arcsim wrinkles

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →