Fall 2016, CS-160 Lecture: Active Shape Modelling (ASM) and Active Appearance Modelling (AAM)

Active Shape Modelling uses a pre-defined template with indexed points to define the outline of a surface.
Active Shape Modelling is useful on faces to detect facial features such as the location of the eyes, nose, mouth, as well as the contour of the face.
This technique was proposed by Tim Cootes and Chris Taylor in 1995.

Here is an active shape model template based on 68 data points. This template needs to be squashed or stretched before it will fit on a person's face.

The Active Shape Model template with 68 data points is matched onto a real face:

Active Appearence Modelling (AAM) is an improvement over Active Shape Modelling.
In AAM, texture on the face is used to compute how the template face should be squashed or stretched to fit the real user's face.
The result is much improved tracking of faces when working with video (i.e. no jitter in index points).

In training, one could use the following dataset to train their own Active Shape Modelling system. The faces in this database contain annotated faces. See the web page at http://personalpages.manchester.ac.uk/staff/timothy.f.cootes/data/xm2vts/xm2vts_markup.html

Active Shape Models - Their Training and Application by T. F. Cootes, D. Cooper, C. J. Taylor, and J. Graham. This article is available electronically from the University of Manchester website at http://personalpages.manchester.ac.uk/staff/timothy.f.cootes/Papers/cootes_cviu95.pdf
An Introduction to Active Shape Models by Tim Cootes. This article is available electronically from http://personalpages.manchester.ac.uk/staff/timothy.f.cootes/papers/asm_overview.pdf
Active Appearence Models by T. F. Cootes, G. J. Edwards, and C. J. Taylor. This article is available electronically from https://www.cs.cmu.edu/~efros/courses/LBMV07/Papers/cootes-eccv-98.pdf