Overview / Usage

In general, all neutral and partial face shape are registered against the ground truth, then the 3D image registration error is measured based on the difference between two images. Unfortunately, such measurement method is unable to justify if the 3D face registration for unconstraint viewpoint is successful. Currently, under such constraint, laborious qualitative analysis take place. So, by classifying unconsraint viewpoint registration using deep learning, it is to replace the overarching qualitative analysis for classification.

Technologies Used

Intel® Optimization for Caffe
Intel's Neural Compute Stick 2
Kniect Xbox 360