Active Appearance Model C++ Library (AAMLibrary)

Dependencies

• opencv 1.0 or later
• cmake 2.6 or later

How to build your program

mkdir build

cd build

cmake ..

make

Quick Tutorial

Prepare:

• For model training, you should have several pairs of images and annotations. AAMLibrary supports pts and asf format.
• Download the imm dataset from AAM-API's homepage link: IMM Dataset
• Download FRANCK dataset annotated by Cootes link: Cootes's Dataset
• Download helen dataset from this link: Helen Dataset
• You can generate your pts file by using Cootes's am_tools and asf file by using Stagmman's aam-api.

Training

• Train the Cootes's basic active appearance models using 3 parymid levels

  ./build -t 0 -p 3 ../helen jpg pts haarcascade_frontalface_alt2.xml basic.amf

• Train the Matthews and S. Baker's Inverse Compositional models using 2 parymid levels

  ./build -t 1 -p 2 ../helen jpg pts haarcascade_frontalface_alt2.xml ic.amf

Fitting

• Image alignment on an image

  ./fit my.amf haarcascade_frontalface_alt2.xml test.jpg

• Face tracking on a video file

  ./fit my.amf haarcascade_frontalface_alt2.xml test.avi

Question

if you have any question, contact me at [email protected], THANKS.