Methods and systems for capturing 3D geometry are becoming increasingly commonplace-and with them a plethora of 3D data.Much of this data is unfortunately corrupted by noise, missing data, occlusions or other outliers.However, when we are interested in the shape of a particular class of objects, such as human faces or bodies, we can use machine learning techniques, applied to clean, registered databases of these shapes, to make sense of raw 3D point clouds or other data.This has applications ranging from virtual change rooms to motion and gait analysis to surgical planning depending on the type of shape.In this chapter, we give an overview of these techniques, a brief review of the literature, and comparative evaluation of two such shape spaces for human faces.