Copyright Notice: The documents distributed here have been provided as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
Presentation at Harvard, November 2007 (also talk at IEEE CVPR 2007, Minneapolis)
Hierarchical Models for 3D Reconstruction Paper (pdf), Talk (pdf), Video1, Video 2 (automatic 3D reconstructions from the movie Run Lola Run)
Presentation at BIRS, November 2006 (also talk at IEEE CVPR 2006, New York)
Learning Joint Top-down and Bottom-up Processes for 3D Visual Inference Paper (pdf), Talk (pdf), Video1 (variable background), Video 2 (3D reconstruction INRIA pedestrian), Video 3 (multiple people, automatic detection and 3D reconstruction)
Presentation at CMU, April 2006 (also talk at IEEE CVPR 2005, San Diego)
Discriminative Density Propagation for Visual Tracking Paper (pdf), Talk (pdf), Video 1 (dance), Video 2 (bend and pick), Video 3 (wash), Video 4 (jump)
Static and Dynamic Ambiguities for 3D Human Pose Reconstruction
Paper: 3D Human Motion Reconstruction in Monocular Video. Techniques and Challenges (C. Sminchisescu), in Human Motion Capture: Modeling, Analysis, Animation, Springer, 2007
Additional Slides (pdf)
Video Set 1: Static, 3D from Monocular, Human Pose Reconstruction Ambiguities (Video)
Video Set 2: Dynamic, 3D from Video, Monocular Human Pose Reconstruction Ambiguities (noisy dynamics): Original Video, 2d tracking hypothesis 1, 3D reconstruction hypothesis 1 (filtered, smoothed), 2d tracking hypothesis 2, 3D reconstruction hypothesis 2 (filtered, smoothed)
Video Set 3: Dynamic, 3D from Video, Monocular Human Pose Reconstruction Ambiguities (state spaces restricted to non-linear manifolds learned from human motion capture data): Original Video, 2d tracking hypothesis 1, 3D reconstruction hypothesis 1, 2d tracking hypothesis 2, 3D reconstruction hypothesis 2
Presentation at the Royal Swedish Academy of Sciences, Mittag-Leffler Institute, March 2003
Optimization methods for ambiguous visual models with applications on 3D human tracking from monocular video given by Cristian Sminchisescu on 12/03/2003 at the Mittag-Leffler Institute of the Royal Swedish Academy of Sciences. The topic of the meeting was Vision from a Mathematical Perspective.
Talk at the IEEE CVPR, June 2003 (Madison, USA)
Kinematic jump sampling methods for monocular 3D human tracking given by Cristian Sminchisescu on 18/06/2003 at CVPR 2003.
Associated Videos