FACE ALIGNMENT WITH DEEP REGRESSION

Abstract

In this paper, we present a deep regression approach for face alignment. The deep regressor is a neural network that consists of a global layer and multistage local layers. The global layer estimates the initial face shape from the whole image, while the following local layers iteratively update the shape with local image observations. Combining standard derivations and numerical approximations, we make all layers able to back propagate error differentials, so that we can apply the standard backpropagation to jointly learn the parameters from all layers. We show that the resulting deep regressor gradually and evenly approaches the true facial landmarks stage by stage, avoiding the tendency that often occurs in the cascaded regression methods and deteriorates the overall performance: yielding early stage regressors with high alignment accuracy gains but later stage regressors with low alignment accuracy gains. Experimental results on standard benchmarks demonstrate that our approach brings significant improvements over previous cascaded regression algorithms.

Existing System

Constraint local models (CLMs), Deep Learning-Based Methods.

Proposed System

In this paper, we propose a deep regression approach that adopts the backpropagation algorithm to jointly optimize a deep structure. The structure consists of two subnetworks: a global layer and multistage local layers. The latter subnetwork is similar to the structure of SDM and each local layer contains a local feature extraction sublayer and a local regressor. The former subnetwork, i.e., the global layer, aims to provide an initial result regressed from the facial image as the input of the latter local regressors. Deep regression differs from cascaded regression in that regressors are jointly learned, instead of learned layer-wisely. In deep regression, regressors are stacked as a deep network and they are jointly optimized by the backpropagation algorithm. The resulting deep regressor gradually and simultaneously reduces the bias and the variance of the estimation from the first regressor to the last regressor, thus yielding a better facial landmark location. This reduces the possibilities that early regressors move landmarks to bad locations, at which later regressors tend to fail to give precise predictions. Experimental results show that the deep regression approach consistently achieves superior or comparable results on several challenging face alignment data sets. The joint learning scheme is quite general and can be applied to a wide range of cascaded regression-based methods, improving their accuracies without harming their computational efficiencies. In this method, we apply the joint learning to the SDM and observe a significant performance improvement.

Architecture

BLOCK DIAGRAM