Synthetic Human Model Dataset for Skeleton Driven Non-rigid Motion Tracking and 3D Reconstruction

We introduce a synthetic human model dataset for evaluating non-rigid 3D human reconstruction based on conventional RGB-D cameras. The dataset consist of seven motion sequences of different actions of a single human model. For each motion sequence per-frame ground truth geometry of the human model and ground truth skeleton are given. The dataset also contains skinning weights of the human model. Our synthetic dataset can be downloaded at: https://doi.org/10.25919/5c495488b0f4e

Contact Dr. Peyman Moghadam for further information.

Detailed documentation and instructions on how to use the data are available in the Synthetic-3D-Reconstruction

One sample data from ’boxing’ sequence is shown. Each frame in the sequence consists of RGB image, depth image, skeleton and ground truth geometry (left to right)

RGB image, depth image, ground truth mesh and 3D skeleton of a particular frame in ‘Boxing’ motion sequence.

Attribution

To attribute this database, please include the following citations:

  1. Elanattil, Shafeeq; Moghadam, Peyman (2019): Synthetic Human Model Dataset for Skeleton Driven Non-rigid Motion Tracking and 3D Reconstruction. v1. CSIRO. Data Collection. https://doi.org/10.25919/5c495488b0f4e
  2. S. Elanattil, P. Moghadam, S. Denman, S. Sridharan and C. Fookes, “Skeleton Driven Non-Rigid Motion Tracking and 3D Reconstruction,” 2018 Digital Image Computing: Techniques and Applications (DICTA), Canberra, Australia, 2018, pp. 1-8. doi: 10.1109/DICTA.2018.8615797

Licensing

The Synthetic Dataset is licensed under CSIRO Data Licence.