Facial landmarks prediction for mobile applications

Face landmarks refer to localization of facial key points. Facial landmarks have been successful applied to head pose estimation, face alignement, face swap applications and more.
We present facial landmark prediction convolutional neural network for mobile applications. We compare neural network in terms of precission, mean error on IBUG, AFLW datasets and also in terms of complexity. Neural network is trained on 300-W and 300-VW datasets. Morover, we present mobile application for facial landmark estimation from front facing camera.

"Algorithms for Image Processing and Analysis" chaired by prof. Władysław Skarbek

Keywords: Neural networks, vision processing, facial landmarks, cnn, landmark estimtation