More Research
DiffAvatar Geometry Transfer Garment3DGen

Garment3DGen: 3D Garment Stylization and Texture Generation

Nikolaos Sarafianos, Tuur Stuyck, Xiaoyu Xiang, Yilei Li, Jovan Popovic, Rakesh Ranjan
Meta Reality Labs

arxiv 2024
Paper Video Code & Data (Coming)

Garment3DGen stylizes the geometry and textures of real and fantastical garments that we can fit on top of parametric bodies and simulate.

3D Garment Generation and Applications





Mesh Deformations



Applications: Hand-Garment interaction in VR





Applications: Sketch to 3D Garment

Qualitative Results





Garment3DGen: Architecture

Given an RGB image we first perform 3D reconstruction using single image to multi-view consistent image generation techniques. We use this output mesh as 3D pseudo-ground-truth to guide a mesh deformation process to deform an input mesh geometry to match the target. Our output geometries preserve the sturcture and topology of the input geometry (as seen by the UV colors of the input and output geometries), contain holes in the neck/arms/waist areas such that they can be fit to bodies and are of good mesh quality to be physically simulated. Finally we generate high-resolution UV texture maps that are faithful to the input image to obtain the final 3D garment asset.

Garment3DGen video (contains sound)

BibTeX

@article{sarafianos2024garment3dgen,
    title={Garment3DGen: 3D Garment Stylization and Texture Generation},
    author={Sarafianos, Nikolaos and Stuyck, Tuur and Xiang, Xiaoyu and Li, Yilei and Popovic, Jovan and Ranjan, Rakesh},
    journal={arXiv preprint arXiv:2403.18816},
    year={2024}
}