3D Morphable Models

Bernhard Egger
Massachusetts Institute of Technology

A 3D Morphable Model (3DMM) is a statistical object model separating shape from appearance variation. Typically, 3DMMs are used as a statistical prior in computer graphics and vision. A model is learned from high-quality 3D scans of multiple object instances. It reduces the dimensionality and provides a low-dimensional, parametric object representation. The resulting model is generative, which means that from a set of randomly sampled parameters a novel realistic object instance arises. The main focus of our research is on the building of 3DMMs and jointly model them with attributes like age, weight or sex and the usage of 3DMMs in an Analysis-by-Synthesis setting, to reconstruct 3D shape and appearance from a still 2D image even under occlusions. This talk will combine a historical perspective, applications in computer vision, synergies with current deep learning based methods as well as an outlook to future directions.

Presentation (PDF File)

Back to Workshop I: Geometric Processing