CLOTH3D: Clothed 3D Humans

TL;DR

CLOTH3D introduces a large-scale synthetic dataset of 3D clothed human sequences with diverse garments and realistic cloth dynamics, along with a generative model for realistic garment synthesis on various poses and shapes.

Contribution

The paper presents the first extensive synthetic dataset of 3D clothed humans and a novel CVAE-based generative model using graph convolutions for realistic garment synthesis.

Findings

Dataset contains diverse garment types, topologies, and fabric properties.

The GCVAE model effectively generates realistic 3D garments.

Cloth dynamics are realistically simulated on various poses and body shapes.

Abstract

This work presents CLOTH3D, the first big scale synthetic dataset of 3D clothed human sequences. CLOTH3D contains a large variability on garment type, topology, shape, size, tightness and fabric. Clothes are simulated on top of thousands of different pose sequences and body shapes, generating realistic cloth dynamics. We provide the dataset with a generative model for cloth generation. We propose a Conditional Variational Auto-Encoder (CVAE) based on graph convolutions (GCVAE) to learn garment latent spaces. This allows for realistic generation of 3D garments on top of SMPL model for any pose and shape.