Discriminative Deep Face Shape Model for Facial Point Detection

摘要: Facial point detection is an active area in computer vision due to its relevance many applications. It a nontrivial task, since facial shapes vary significantly with expressions, poses or occlusion. In this paper, we address problem by proposing discriminative deep face shape model that constructed based on augmented factorized three-way Restricted Boltzmann Machines model. Specifically, the combines top-down information from embedded patterns and bottom up measurements local detectors unified framework. addition, along model, effective algorithms are proposed perform learning infer true locations their measurements. Based 68 points detected images both controlled "in-the-wild" conditions. Experiments benchmark data sets show effectiveness of algorithm against state-of-the-art methods.