GTA-Human II Dataset

[Go back to Home]

This is the latest version of our datasets, and is built upon GTA-V for expressive human pose and shape estimation. It features multi-person scenes with SMPL-X annotations. In addition to color image sequences, 3D bounding boxes and cropped point clouds (generated from synthetic depth images) are also provided. Please contact Zhongang Cai (caiz0023@e.ntu.edu.sg) for feedback.

Download links and a toolbox can be found here.

Citation

      @ARTICLE{10652891,
          author={Cai, Zhongang and Zhang, Mingyuan and Ren, Jiawei and Wei, Chen and Ren, Daxuan and 
                  Lin, Zhengyu and Zhao, Haiyu and Yang, Lei and Loy, Chen Change and Liu, Ziwei},
          journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
          title={Playing for 3D Human Recovery}, 
          year={2024},
          volume={},
          number={},
          pages={1-12},
          keywords={Three-dimensional displays;Annotations;Synthetic data;Shape;Training;
                  Parametric statistics;Solid modeling;Human Pose and Shape Estimation;
                  3D Human Recovery;Parametric Humans;Synthetic Data;Dataset},
          doi={10.1109/TPAMI.2024.3450537}
      }