FBK home > INFORMATION TECHNOLOGY > Technologies of Vision > Databases > DPOSE

DPOSE - multi-view dataset with head pose ground truth

about DPOSE

* DPOSE contains sequences acquired from 16 subjects, where the subject is either (i) rotating in-place at the room center, or (ii) moving around freely in a room, and moving their head in all possible directions. The dataset consists of over 50000 images, recorded from four cameras with overlapping field of view. Images have resolution 1024x768 and are stored in jpeg format. Head pan, tilt and roll measurements for various poses are recorded using an accelerometer, gyro, magnetometer platform strapped onto the head using an elastic band running down from the back of the head to the chin.

Head pan = 270° Head pan = 0° Head pan = 0°
Multi-view samples of DPOSE, with ground truth overlay (face crops non provided with the dataset). From left to right: Head pan = 270°, 0°, 0°.

* DPOSE comes with camera calibration files: for every camera, we provide all the information needed to compute the image projection (pixel coordinates) of a 3D point. C++ code implementing this projection is provided.

room
Room dimensions, ground coordinates, and camera positions

* We also provide tracking of the subjects (ground coordinates), computed with a color based particle filter.

download DPOSE

DPOSE is provided for research or academic purposes only. If you use this dataset please cite our ACCV'12 paper referenced at the bottom of this page.

DPOSE is large (total 50 GB): it comes with 18 files of about 3 GB each (tar archives, one per subject, plus recordings of the empty room), each of which can be downloaded and used independently. Every archive contains three synchronized image sequences with head pose references.


Please provide your contact information. After registration you will receive an e-mail notification with download instructions.
your name
email (required)
affiliation
Your contact information will only be used for notifications regarding this dataset.

related publications

* A. Rajagopal, R. Subramanian, R. Vieriu, E. Ricci, O. Lanz, N. Sebe, K. Ramakrishnan
An Adaptation Framework for Head Pose Estimation in Dynamic Multi-view Scenarios
Asian Conference on Computer Vision - ACCV 2012, Daejeon, Korea, November 5-9, 2012 [mpg video]

* Y. Yan, R. Subramanian, O. Lanz, N. Sebe
Active Transfer Learning for Multi-view Head-pose Classification
International Conference on Pattern Recognition - ICPR 2012, Tsukuba, Japan, November 11-15, 2012

* R.L. Vieriu, A. Rajagopal, R. Subramanian, O. Lanz, E. Ricci, N. Sebe, K. Ramakrishnan
Boosting-based Transfer Learning for Multi-View Head-Pose Classification from Surveillance Videos
European Signal Processing Conference - EUSIPCO 2012, Bucharest, Romania, August 27-31, 2012, pp. 649-653

info

Research in collaboration with M-HUG Lab, DISI-UNITN. For further information contact Oswald Lanz, e-mail l a n z (at) fbk. eu