Introduction
We have created a public database of videos for head tracking and pose estimation. The database is a synthetic replica of the UPNA Head Pose Database. Twelve videos per user have been thus generated, which include 6 guided-movement sequences and 6 freemovement sequences. The videos have been generated with a 1280×720 pixel resolution, at 30 frames per second and are provided in MPEG-4 format. Translation and rotation axes are the same as in UPNA Head Pose Database and are defined in the figure below.
Definition of translation and rotation axes
Each video is associated with three text files. One contains the head pose with respect to the camera (same as those in UPNA Head Pose Database). The others cointain, the 2D projections (in pixels) of 43 annotated 3D facial points, what we will call the 2D ground truth landmarks, and the bounding box with the following format: minX minY width height. Translations are given in millimeters and rotations in degrees.
Sample images taken from UPNA Head Pose Database and Synthetic UPNA Head Pose Database:
References
- Andoni Larumbe, Mikel Ariz, José J. Bengoechea, Rubén Segura, Rafael Cabeza, Arantxa Villanueva, Improved strategies for HPE employing learning-by-synthesis approaches, Fifth International Workshop on Assistive Computer Vision and Robotics (ACVR), International Conference on Computer Vision (ICCV ’17)
- Mikel Ariz, José J. Bengoechea, Arantxa Villanueva, Rafael Cabeza, A novel 2D/3D database with automatic face annotation for head tracking and pose estimation, Computer Vision and Image Understanding, Volume 148, July 2016, Pages 201-210, ISSN 1077-3142
Download the database
This database is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. The data are only to be used for non-commercial scientific purposes. If you use this dataset in scientific publication, please cite the aforementioned papers.
You can download ...