PersEmoN: A Deep Network for Joint Analysis of Apparent Personality, Emotion and Their Relationship

Personality and emotion are both central to affective computing. Existing works solve them individually. In this paper we investigate if such high-level affect traits and their relationship can be jointly learned from face images in the wild. To this end, we introduce PersEmoN, an end-to-end trainable and deep Siamese-like network which we call emotion network and personality network, respectively. It consists of two convolutional network branches, one for emotion and the other for apparent personality. Both networks share their bottom feature extraction module and are optimized within a multi-task learning framework. Emotion and personality networks are dedicated to their own annotated dataset. An adversarial-like loss function is further employed to promote representation coherence among heterogeneous dataset sources. Based on this, the emotion-to-personality relationship is also well explored. Extensive experiments are provided to demonstrate the effectiveness of PersEmoN.