Learning deep representation from coarse to fine for face alignment
This work addresses face alignment for computer vision applications, presenting an incremental improvement over existing methods.
The paper tackles face alignment by training a deep convolutional network from coarse to fine, dividing landmarks into subsets with adjusted weights to learn an initial model and smoothly search for optimal solutions, achieving a 6.33% mean error on the COFW dataset, a 21.37% reduction from the previous best.
In this paper, we propose a novel face alignment method that trains deep convolutional network from coarse to fine. It divides given landmarks into principal subset and elaborate subset. We firstly keep a large weight for principal subset to make our network primarily predict their locations while slightly take elaborate subset into account. Next the weight of principal subset is gradually decreased until two subsets have equivalent weights. This process contributes to learn a good initial model and search the optimal model smoothly to avoid missing fairly good intermediate models in subsequent procedures. On the challenging COFW dataset [1], our method achieves 6.33% mean error with a reduction of 21.37% compared with the best previous result [2].