Automated Emotion Recognition Through Graphical Cues on Comics at Character Scale.

Emotions are psychological reactions to external events. Characters represented in artistic works may manifest emotions in order to replicate credible and human-like behaviors in specific situations. They also provides important hints to better understand the stakes and the tone of story. In comics, markers of emotions can be found in the dialogues or through visual cues specifically drawn by the artists. While automated emotion extraction on textual information is an active research field, few works have addressed this topic through the graphical grammar of comics (and more generally on drawings). In this paper, we propose to review the different visual tools used by artists to convey expressiveness to their characters and how they can be exploited for automated processing. Some of those cues are strongly related to the human body, its representation and mechanisms. Consequently, we propose to study developed methods for those topics on photography or captured videos. Then, we suggest contributions that aimed at facilitating the transition between real and drawn domains.