Skip to content

Point Distribution Model

Tadas Baltrusaitis edited this page Aug 3, 2017 · 2 revisions

Outline

The Point Distribution Model (PDM) is a way to parametrise the shape of a face using a limited set of parameters. In our case the PDM is made up of two terms, rigid and non-rigid shape parameters. Rigid shape parameters describe the placement of the face in the image --- scaling, rotation, and translation. Non-rigid shape parameters on the other hand describe the deformation of the face due to identity or expression, e.g. wider or taller faces, smiles, blinks etc.

Rigid shape parameters

Scaling, translation, and rotation

Non-rigid shape parameters

PDM training

Model visualization

Clone this wiki locally