V-Human dataset

19th August 2022
V-Human dataset

Clothed human shape estimation has been a topic of interest in the computer vision community for many years and recently has become even more relevant due to the leap in performance given by recent deep learning approaches. In order to alleviate the cost and quality issues with existing commercial and freely available datasets respectively, we have created a dataset from synthetic models suitable for training deep learning algorithms for 3D human reconstruction from images. To prepare the dataset we used the fully rigged avatars from the Microsoft Rocketbox library [Gonzalez-Franco et al., 2020], re-targeted Mixamo animations to them, and finally refined them to make them suitable to be used as direct input for learning techniques.

Dataset details

This dataset contains 2160 models, divided into two subsets: training (1944) and test (216).

It contains 90 different subjects (81 for training and 9 for testing) with 24 poses per subject.

For each model, a mesh with its correspondent texture is given in obj format.

Download

You can download the dataset from here.

Publication

Please, if you use this dataset consider citing our work:

J. Gonzalez-Escribano, S. Ruano, A. Swaminathan, D. Smyth, A. Smolic

Texture improvement for human shape estimation from a single image.

Irish Machine Vision and Image Processing Conference, 2022

References

[Gonzalez-Franco et al., 2020] Gonzalez-Franco, M., Ofek, E., Pan, Y., Antley, A., Steed, A., Spanlang, B., Maselli, A., Banakou, D., Pelechano Gómez, N., Orts-Escolano, S., et al. (2020). The rocketbox library and the utility of freely available rigged avatars. Frontiers in virtual reality, 1(article 561558):1–23.