GPI Seminar Series: Dalila Sánchez

Dalila Sánchez Escobedo, Statistical 3D face reconstruction from sparse information,
(rescheduled) Wednesday January 29th, 11:00 - 12:00, Seminar Room D5-007

Statistical 3D face reconstruction from sparse information

Abstract:
In this talk the problem of linearly approximating 3D shape from 2D information in the context of facial analysis will be addressed. In other words, through statistical models we aim to relate 2D and 3D information allowing us to estimate a 3D face structure from 2D information. Different estimation methodologies that incorporate one or more statistical models, as fitting and regression models, will be presented along with their advantages and disadvantages. Also, the study of different data representation for the 2D and 3D information will be aboard. However, to obtained an accurate 3D face estimation, the relationship between 2D and 3D information has to be strong. Experimental tests show that it is possible to strengthen the relationship between 2D and 3D information using different data representation.