Highlights

Human motion analysis in images and video is a central computer vision problem. Yet, there are no studies that reveal how humans perceive other people in images and how accurate they are. We aim to unveil some of the processing as well as the levels of accuracy involved in the 3D perception of people from images by assessing the human performance.

3D Human Motion Capture Systems

  • 200Hz sample rate

  • 10 high speed motion camers

SMI Head Mounted Eyetracking System

  • 200Hz eyetracking sample rate

  • 25fps scene video data

Data Download

In order to have acces to the data you will have to create an account, login and accept the license.

References

The dataset, the analysis on human 3D pose re-enactment, and related experiments are described in detail in:

  • Elisabeta Marinoiu, Dragos Papava and Cristian Sminchisescu, Pictorial Human Spaces: How Well do Humans Perceive a 3D Articulated Pose?, International Conference on Computer Vision (ICCV), 2013 [pdf][bibtex]

  • Elisabeta Marinoiu, Dragos Papava and Cristian Sminchisescu, Pictorial Human Spaces: A Computational Study on the Human perception of 3D Articulated Poses, International Journal of Computer Vision (IJCV), 2016 [pdf][bibtex]

The license agreement for data usage implies the citation of the two papers above. Please notice that citing the dataset URL instead of the publications would not be compliant with this license agreement.

General Setup

Our idea was to show subjects images of humans in different poses (taken from the Human3.6M dataset ), and then asked them to reproduce the poses seen as well as they could. The subjects were explicitly instructed not to mirror the pose, but to reproduce the left and right sides accordingly. The images were projected on a 1.2 meters tall screen located 2.5--3 meters away. The eye tracker calibration was done by asking the subject to look at specific points, while the system was recording pupil positions at each point. The calibration points were projected on the same screen used to project images. For mocap we use the standard calibration procedure.

Subjects

We first analyze the re-enactment performance of 10 subjects, 5 male and 5 female, who did not have a medical history of eye problems or mobility impediments. Moreover, their profession did not require above average neuro-motor skills (as required in the case of dancing, acting or practicing a particular sport). This group is reffered to as the regular subjects. In addition to this, we also analyze the performance of another 4 subjects, 2 males and 2 females, who were all final year choreography students, focusing on modern and classical ballet. This group is referred to as the skilled subjects. All the participants were recruited through an agency and had no link with computer vision.

Data

Please check the following README file for details about data format.

Use the links below to download our data:

Acknowledgements

This work was supported in part by CNCS- UEFISCDI under PCE-2011-3-0438, and JRP-RO-FR-2014-16.