Companion data set to CVPR 2009 paper

Description:

• The data set involves 12 sequences of varying length (2360 multi-view frames in total) for benchmarking 3D human pose estimation systems.
• The data set was recorded outdoors and contains cluttered dynamic backgrounds and lighting changes. The sequences contain diverse actions such as walking, standing, gesticulation and interaction with other persons.
• Images were recorded with three overlapping hardware-synchronized color CCD cameras at 20 fps and VGA resolution. We supply images where lens distortion has already been removed.
• 3D marker data for 17 body locations is made available for all sequences, derived from human labeling (one subject per sequence).

License Terms

This data set is made available to the scientific community solely for non-commercial research purposes such as academic research, teaching, scientific publications, or personal experimentation. Re-distribution is not permitted without the express permission of the authors.

The data set comes "AS IS", without express or implied warranty. Although every effort has been made to ensure accuracy, the University of Amsterdam does not accept any responsibility for errors or omissions.

The use of the data set is to be acknowledged by citing: M. Hofmann and D. M. Gavrila, “Multi-view 3D Human Pose Estimation combining Single-frame Recovery, Temporal Integration and Model Adaptation”, CVPR 2009.