There is a disconnect between what computer vision (and AI) researchers think emotions are and how they are conveyed through facial expressions and body pose, and what the actual science tells us.
Take faces as an example. Undoubtedly, faces offer information that helps us navigate our social world, influence whom we love, and determine who we trust or who we believe to be guilty of a crime. But to what extent does an individual’s face reveal the person’s internal emotions? To what extent (and how) can we design computer vision systems to accurately interpret an emotion or intention from a raised eyebrow, a curled lip, or a narrowed eye? And to what degree are these visual cues influenced by body pose and context?
Recent research shows that faces or body expressions alone are insufficient to perform a reverse inference of image to emotion, and that context, personal believes, and cultural must be accounted for. This workshop will present these limitations and examine several alternative approaches to successfully interpret the emotion and intent of others.