Section: New Software and Platforms

Mixing Body-Part Sequences for Human Pose Estimation

Participants : Cherian Anoop, Mairal Julien, Alahari Karteek, Schmid Cordelia.

The code corresponding to the publication [11] has been released as an open-source MATLAB package along with a dataset for human pose estimation in videos called “Poses in the Wild”. It is available at http://lear.inrialpes.fr/research/posesinthewild/#dataset . This dataset has 30 video sequences generated from three Hollywood movies, namely “Forrest Gump”, “The Terminal”, and “Cast Away”. Each sequence has approximately 30 frames and is manually annotated for human upper-body keypoints, namely (i) neck, (ii) left and right shoulders, (iii) left and right elbows, (iv) left and right wrists, and (v) mid-torso. In comparison to earlier evaluation datasets publicly available for this problem, Poses in the Wild is significantly more representative of real-world scenarios with background clutter, body-part occlusions, and severe camera motion.