ICIAP 2013 Tutorial on
Non-Rigid 3D Reconstruction from Images10 September 2013, 17th International Conference on Image Analysis and Processing (ICIAP 2013), Napoli, Italy |
This tutorial presents an overview of modern computer vision techniques to deal with the problems of 3D reconstruction of deformable shapes from images. The aim is to present principled, practical and robust algorithms to deal with images of objects whose shapes can change through time such as in the cases of human motion analysis, medical imaging and video-surveillance. This tutorial will discuss techniques which are general in their formulation but can be customized to solve different specific problems.Recovering the 3D shape of a deformable surface from a single video stream from image correspondences (a problem denoted non-rigid structure from motion or NRSfM) remains an open problem due to its inherently ill-posed nature — many different 3D shapes can have the same image projections. To resolve these inherent ambiguities, prior knowledge of either the shape of the object or the nature of the deformations must be used in formulating the problem. In this tutorial we will focus on the description of different types of priors which have been proposed in the literature to make the problem tractable. I will also categorize methods according to the optimization strategy used to solve the reconstruction problem. Real examples on sequences displaying human motion and surface deformations such as those in paper and cloth will show the effectiveness of the approaches in dealing with different types of deforming shapes. |
Schedule and Slides – September 10, 2013Download PDF slides HERE
|
Previous editions of the tutorial:
- First edition of the tutorial (full-day, @ISR, Lisbon, april 2009) |
Alessio,
Thanks for sharing the materials. Very useful to learn. Great job!
Hey,
the “Alternative models” pdf is not found (err 404), someone can find it?
Hi Julien, we fixed the broken link, try it now please.
[...] Tutorial [...]
[...] Tutorial [...]