fnsmile is an OpenCV based smile detector which uses, besides a cascade smile model, a face model and a nose model to reduce false alarms as much as possible.

This code is supplied as a demo for this blog article. To summarize the main idea in a couple of words, here are the steps of the algorithm:

• Detect all faces on a picture
• Try to detect a nose on each face. Discard faces with no nose (reduces the face model's false alarms)
• Use the lower part of the face, starting from the center of the nose, to detect a smile

OpenCV's cascade models are usually located in /usr/local/share/OpenCV/haarcascades/. You will most certainly have to append the name of the model to this path.

python fnsmile.py -s haarcascade_smile.xml

python fnsmile.py -s haarcascade_smile.xml -f haarcascade_frontalface_default.xml

python fnsmile.py -s haarcascade_smile.xml -f haarcascade_frontalface_default.xml -n haarcascade_mcs_nose.xml

-v 1 

-v filename

Amine Sehili ([email protected])

This program is available under the GNU GENERAL PUBLIC LICENSE Version 3.