Deep Fashion3D V2: Release Note

By Heming Zhu

Overview

We present a major update upon the original version of Deep Fashion 3D dataset . Compared with the previous version of Deep Fashion3D dataset, Deep Fashion3D V2 is futher equipped with: (1) detailed registered garment meshes with category-specific triangulation, (2) high-resolution texture maps (2048 X 2048 px), (3) more precise and accurate feature line annotations, and (4) garment SMPL pose parameters. We believe this update will further boost the research in high-quality, real-world garment modeling and generation.

Download

Please check out the following links for downloading Deep Fashion3D V2 and the paper.

Data Samples

In below, we show some data samples ( detailed registered garment meshes with category-specific triangulation) from Deep Fashion3D V2. You may press 'g' to switch on/off the texture, and 'r' to reset the view point. The code for visualizing the meshes is partial ly borrowed from Yuliang Xiu's implementation for ECON.

Assets and Annotations

High resolution textures

In below, we show the high-resolution textures (2048px x 2048px) for the registered garment meshes. Note that the textures are aligned within each garment category.

Precise feature line annotations

In the previous version of Deep Fashion 3D dataset , we proposed to annoate the garment feature lines as point cloud bands near the corresponded regions on the garment points clouds. Although the feature lines can be served as prominent features for deforming the garment meshes, they may lead to zig-zags and inaccuracies due the thickness of the annotated feature line point clouds. To address this issue, in this major update, we provide the precise and accurate feature line annotations which exactly indicates the boundaries of the garment point clouds.

Garment pose annotations

As is shown in the figure below, for each garment point cloud in Deep Fashion3D V2, we provide SMPL pose parameters, scale, and translation that can be adopted for aligning the SMPL body mesh to the garment point clouds.

Citation

If you use Deep Fashion3D or Deep Fashion3D V2 in your work, please consider citing our paper, thanks!

@inproceedings{zhu2020deep,
    title={Deep Fashion3D: A Dataset and Benchmark for 3D Garment Reconstruction from Single Images}, 
    booktitle={Computer Vision -- ECCV 2020},
    author={Heming, Zhu and Yu, Cao and Hang, Jin and Weikai, Chen and Dong, Du and Zhangye, Wang and Shuguang, Cui and Xiaoguang, Han},
    year={2020},
    publisher={Springer International Publishing},
    pages={512--530},
    isbn={978-3-030-58452-8}
}

Heming Zhu