IJCV 2021 – Human retargeting dataset

This dataset is composed of paired motion sequences from different characters and annotated motion retargeting constraints. For each video sequence, we provide a refined 3D actor reconstructed motion and the actor
body shape estimated with Alldieck et al. (2018).

@Article{Gomes2021,
author={Gomes, Thiago L. and Martins, Renato and Ferreira, Jo{\~a}o and Azevedo, Rafael and Torres, Guilherme and Nascimento, Erickson R.},
title={A Shape-Aware Retargeting Approach to Transfer Human Motion and Appearance in Monocular Videos},
journal={International Journal of Computer Vision},
year={2021},
month={Apr},
day={29},
issn={1573-1405},
doi={10.1007/s11263-021-01471-x},
url={https://doi.org/10.1007/s11263-021-01471-x}
}

The dataset sequences are divided into two sets: training and test.

Training Set

For each actor we provide 4 types of data: 4-min videos; 8-views; A-pose and dance.

• 4-min: 4 minutes of random movements performed by each actor. The directory contains a folder called images with the .jpg images, a text file (person.txt) with the heigth and smpl parameters, and a consensus of the smpl shape (.pkl).
• 8-views: This folder contains 8-views from each actor with 45 degree variation between each pose. It has three additional folders: segmentations with binary masks of the actors; semantic_label containing the human body parts labeled semantically; and smpl_pose with the SPIN pose estimations.
• A-pose: Original 360 degree movements used to build the 8-views folder.
• dance: Random dance movements performed by each actor.

Test Set

For the test set, we provide 8 movements performed by each actor:

• • box
  • cone
  • fusion dance
  • jump
  • pull down
  • shake hands
  • spinning
  • walk

Each movement folder contains a text file (person.txt) with the height and SPML parameters, as well as the annotated 2D and 3D motion restrictions. There is also three sub-folders containing: i) images from the scene; ii) openpose with the actors 2D poses from OpenPose/SPIN; and iii) smpl_pose with the corresponding SMPL model estimation.

Copyright Notice

This dataset is published under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) License. This means you must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. You may not use the material for commercial purposes.