Description

Recovering deformable 3D motion from temporal 2D point tracks in a monocular video is an open problem with many everyday applications throughout science and industry, or augmented reality [1]. Recently, several techniques have been proposed to simultaneously obtain camera motion and non-rigid shape [2,3,4], but they can exhibit poor reconstruction performance on complex motion. In this project, we will analyze these situations for primitive human actions such as walk, run, sit, jump, etc on different scenarios [5,6]. To this end, we first review current techniques to finally present a novel method where different types of priors are exploited. This master thesis will be carried on in the Institut de Robòtica i Informàtica Industrial, at the Universitat Politècnica de Catalunya (under direction of Dr. Antonio Agudo and Dr. Francesc Moreno).


Requisites:

Candidates with a background in mathematics, computer vision and good programming skills (Matlab/C++) are particularly encouraged to apply.


For additional information, please contact Dr. Antonio Agudo at aagudo@iri.upc.edu


[1] J. Pilet, V. Lepetit and P. Fua. Fast Non-Rigid Surface Detection, Registration and Realistic Augmentation. In IJCV 76(2), 109-122, 2007.

[2] P.F.U. Gotardo and A. M. Martinez. Kernel non-rigid structure from motion with complementary rank-3 spaces. In ICCV, 2011.

[3] Y. Dai, H. Li and M. He. A simple prior-free method for non-rigid structure from motion factorization. In CVPR, 2012.

[4] M. Lee, J. Cho, C. H. Choi and S. Oh. Procrustean normal distribution for non-rigid structure from motion. In CVPR, 2013.

[5] Carnegie Mellon University Motion Capture Database.

[6] HMDB: A Large Human Motion Database.

The work is under the scope of the following projects:

• RobInstruct: Instructing robots using natural communication skills (web)