Learning Local Objective Functions for Robust Face Model Fitting

Abstract

Model-based techniques have proven to be successful in interpreting the large amount of information contained in images. Associated fitting algorithms search for the global optimum of an objective function, which should correspond to the best model fit in a given image. Although fitting algorithms have been the subject of intensive research and evaluation, the objective function is usually designed ad hoc, based on implicit and domain-dependent knowledge. In this article, we address the root of the problem by learning more robust objective functions. First, we formulate a set of desirable properties for objective functions and give a concrete example function that has these properties. Then, we propose a novel approach that learns an objective function from training data generated by manual image annotations and this ideal objective function. In this approach, critical decisions such as feature selection are automated, and the remaining manual steps hardly require domain-dependent knowledge. Furthermore, an extensive empirical evaluation demonstrates that the obtained objective functions yield more robustness. Learned objective functions enable fitting algorithms to determine the best model fit more accurately than with designed objective functions.

Generality and legibility in mobile manipulation

“…These hulls are similar to Workspace Goal Regions (Berenson et al 2009), except that our hulls refer to base positions, whereas in Berenson For each of the 12 different cup positions, we get one hull. In the next step, we compile all hulls into a generalized compact representation using a Point Distribution Model (PDM), which is a well established method in the field of face recognition (Wimmer et al 2008). As input a PDM requires n points that are distributed over the contour.…”

Multi-Kernel Appearance Model

“…One can try to measure the resemblance between each pixel in the region of interest and the point we are seeking. [52] proposes a tree-based regression algorithm over patches characterized by haar-like features. Other methods prefer to estimate the displacement vector v that continuously relates a patch location to the target point.…”

Towards Unconstrained Face Recognition Using 3D Face Model

“…Further, this approach is less laborious because the objective function design is replaced with automated learning. For details we refer to Wimmer et al (2008). The geometry of the model is controlled by a set of action units and animation units.…”