scoobadood / faceoff

This is an implementation of Michael Zollhöfer's 'Reconstruction of Personalized Avatars' work, recreating an input face from a Kinect scan using a 3D morphable map

FaceOff

This is an attempt to recreate Michael Zolloeffer's work from Part I of his paper Zollhöfer, M. 2015, 'Real-Time Reconstruction of Static and Dynamic Scenes', pp. 1–196.

The paper describes a method for constructing personal avatars from Kinect RGB-D scans using a 3D Morphable Face Model It's also an attempt by me to get to grips with some fundamental CV toolkits including OpenCV and PCL

Dependencies

FaceOff uses OpenCV (built with version 2.4) for manipualting data and OpenNI for capturing Kinect input RGB-D frames.

Building

Build using CMake

Running

Currently launching the FaceOff binary will open a window and display the captured RGB image. If there's a face present (occupying at least 30% of the width/height of the display) it will be outlined with a blue box and left and right eyes will be identified with green boxes.