This manuscript presents the author’s most significant works conducted from 2011 to 2017 on the topic of processing and analyzing 3D geometric data, and in particular deformations of 3D human shapes and their accessories. Applications of such 3D data include product design, avatar creation, and recognition tasks. Inferring high-level information from raw scan data is challenging due to high levels of noise in the captured data and a high degree of variability in geometry both across different human subjects and across different poses. For this reason, to date, most of the 3D scan data used in applications are processed with the help of manual input. The goal of our work can be summarized as providing automatic methods for processing and analyzing raw geometric 3D data showing the shape and deformations of humans and their accessories. To achieve this goal, we take advantage of the fact that humans and their clothing allow for a limited set of deformations, and use deformation models to constrain the space of possible human 3D shapes. This document presents contributions based on three deformation models. The first part considers a near-isometric deformation model defined on partial regions that can be used to model loco-motions of humans as well as cloth deformations. The main novelty of this model is that it allows for processing that is robust to acquisition noise. The second and third parts consider generative models for human body and face shape that are learned from a training database of 3D scans. For both bodies and faces, we present automatic processing pipelines that allow to build generative models based on databases of thousands of raw scans. We further present applications of both the near-isometric deformation model and the generative model of human body shape.