##Architecture FaceViewWithAnalysis calls all the CV functions on frames inside FaceViewWithAnalysis.processImage(). It performs consistency analysis on every frame and performs face prediction every few seconds.
FacePredictor takes images as input, determines where the face is located and guesses who the face belongs to.
BackgroundConsistencyAnalysis is used to determine whether the face is actually a foreground face, or likely a picture of a face (ie, a spoofed face).
All computations are done inside AsyncTasks, inside (UI) Activitys. For simplicity, no Service is used. Therefore, if you background the app while computations are taking place you may see crashes (I haven't seen any, but I wouldn't be surprised).
##Setup on Mac OS X Setup the Android SDK with Eclipse as per Google's instructions. I don' think there should be any dependence on the Android NDK.
Import the application directory into an Eclipse workspace. There should be no additional setup.
##Current Status Main Components
- One component is capable of finding faces on the screen in real time. It is a little slow, since I'm no longer using the C++ code.
- Another component recognizes faces from a training set. I've setup a very small training set (inside src).
The two above components are not yet connected.
##Proposal
https://docs.google.com/document/d/1qCn1KuC45r-uz-HTVkwYc8mqkEIqjNAC2T6LP0xs_6E/pub
##Testing Source code for a test project are contained inside the FacePreview_Test sub-project.