Deep Regression for Face Alignment

TL;DR

This paper introduces a deep regression method for face alignment using a multi-stage architecture and dropout optimization, achieving state-of-the-art accuracy in facial landmark detection.

Contribution

It proposes a novel deep regression framework with a global and multi-stage local layers, optimized via back-propagation and dropout, improving face alignment accuracy.

Findings

Achieves state-of-the-art face alignment results

Gradually approaches true facial landmarks stage by stage

Avoids over-strong early regressors and over-weak later regressors

Abstract

In this paper, we present a deep regression approach for face alignment. The deep architecture consists of a global layer and multi-stage local layers. We apply the back-propagation algorithm with the dropout strategy to jointly optimize the regression parameters. We show that the resulting deep regressor gradually and evenly approaches the true facial landmarks stage by stage, avoiding the tendency to yield over-strong early stage regressors while over-weak later stage regressors. Experimental results show that our approach achieves the state-of-the-art