-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Written in C++ and GLSL. Currently attempts to detect and recognise faces and emotion.
Goto Mission & Goals
Currently almost finished.
Uses Viola-Jones' paper as the main concept. Can create Haar Features for accurate object detection.
Currently being worked on.
Uses Facebook's Deepface paper as inspiration.
2D fiducial landmarking points given by OpenFace (required external software).
Facial alignment & representation & verification is done with my software using GLSL programming to take advantage of GPU processing.
Uses Deep Learning and Convolutional Neural Networks for representation.
Currently not finished.
Identification is done by repeatedly verifying faces.
My program's workflow: detect, align, represent, classify, verification, identification.
(strikethrough was Facebook's Deepface method, for higher accuracy although more difficult to implement)
Using OpenFace’s action units (similar to the aforementioned Haar face classifiers), emotion can be extrapolated from a face. This is not computed by my program but by OpenFace and is not my work.
Goto Technical Documentation & my thoughts logs for more indepth writing by me
libraries & software required -
- GLFW
- GLEW
- libPNG
- OpenFace