In recent years, deep neural networks have been widely concerned by researchers in facial expression recognition. However, insufficient facial training data of the public available database is a major challenge in deep learning, which will lead to an obvious decrease in the effectiveness of learning resu many data augmentation techniques have thus been widely used to enrich the training dataset. In this paper, we introduce a contextual loss function to construct a Contextual Generation Adversarial Network with one generator and one discriminator. The proposed method can map the neutral expression to six basic expressions to expand the database. The experimental results on CK + and KDEF databases show that the proposed method can effectively improve the ability to extract facial features and the ability to generate higher quality images. The data augmentation used the proposed method improves the recognition rate of facial expressions on KDEF and CK + datasets.
展开▼
