Negin, Fırat Özdemir, Ceyhun Burak Akgül, Kamer Ali Yüksel, Aytül Erçil
Organizations: Sabancı University VPA Lab (SU-VPA), Vistek ISRA Vision (VIV)
Agreement: Any resulting publication which will use or refer to the database, should have a reference to
the database, VIPSAFE Project, Sabanci University and VPALAB.
The minimal dataset may be freely downloaded from here (84 MB).
The full database (142 GB) may be obtained by application from the contact person above.
The WorkoutSU-10 exercise dataset comprises a collection of sequences of human body movements represented by 3D positions of skeletal
joints. The dataset was collected at Sabancı University in Istanbul as part of the research project ViPSafe supported
partially by TÜBITAK (agreement 109E134).
If you use this dataset, the following paper should be cited in any
F. Negin, F. Özdemir, C. B. Akgül,
K. A. Yüksel, A. Erçil
“A Decision Forest Based Feature
Selection Framework for Action Recognition from RGB-Depth Cameras.”
Special Session on Recent Advances
on RGB-D Camera Applications
International Conference on Image
Analysis and Recognition (ICIAR 2013)
The dataset comprises of 1500 sequences in total, collected from 15
people performing 10 different exercises. The motion files contain tracks of 20
joints estimated using the MS Kinect SDK. The body pose is captured at a sample
rate of 30Hz.
The participants were recruited at the Computer Vision Laboratory at
Sabancı University (VPALAB). All participants filled a consent form. Each
recording session has taken 50 to 90 minutes. Some of the participants were already
familiar with the domain of machine learning and computer vision. Relevant
statistics of the participants are listed below:
153-191cm tall with an
average height of 179 cm,
20 to 30 years old with an
average of 24 years of age.
For a full description please visit WorkoutSU-10 homepage