#face detect and localization My demo for face detection and facial point localization. I reimplement the algorithm in DLIB, it now depends on OpenCV, Boost(training phase only) only, you may find it useful in some case.

you can choose FHOG or OpenCV's haar detector. FHOG has a better ROC curve but takes a lot of time though... facial point localization takes 4-5 ms on my computer, so it's pretty fast..

##demo This is a demo image for the algorithm( 68 facial points, trained on Helen dataset), the localization part is fast and accurate enough in most situations ....

##how to use checkout and compile( CMake is the easist way, since it depends only on OpenCV in test time, you can easily set it up on Windows too)

1. FHOG + facial point localization

see shape_predictor/test_main.cpp for example, download these two models: fhog detector and facial_point_for_fhog, put them in the right folder, you are good to go.

2. Haar + facial point localization

see shape_predictor/test_haar_main.cpp, download haar_detector and facial_point_for_haar

note these 2 facial point models are not the same!

##and ... This also contains opencv implementation for channels feature and soft cascade decision tree, we use it for pedestrian detection, it works well! but I forget where I put the trained model ...