Cascaded Nonlinear Shape Model Regression

摘要: This paper targets deformable face model matching in images using cascaded regression techniques. Recently, the strategies have became rather popular solutions to solve nonlinear objective functions by learning a pipeline sequence of linear regressors. However, despite their success, standard formulation do not enforce shape consistency through cascade (mostly because individual regressors are learnt independently). In this we revisit framework and propose number improvements. First explore simplicity compactness for such tasks, effectively solving previous drawback. Then extend module into version, means recent Convolutional Neutral Networks (CNNs) techniques, modified include weighted aware loss function. Moreover, since CNNs often require massive amounts data perform well, took advantage probabilistic efficiently bootstrap new data. Our method is evaluated several databases (LFPW, LFW, HELEN 300W), where results demonstrate effectiveness proposed method.