Contenuto scaricabileScarica il pdf
Emotion classification in images
Our goal in this research is to perform semantic extraction from images. The semantics in question can range from objects, colors, seasons, or events in an image; the topic is well researched in the field of Computer Vision but only recently have researchers begun to explore complex semantics: ideas that build onto simplistic semantics like colors, contours etc. In our research, we would like to extract the emotion an image evokes in the viewer's mind. We would follow the definition of emotion as described by Parrot's Hierarchy and train a Deep Neural Network model to perform the task. Additionally, we would like to define, and explore other complex semantics like 'masculinity', 'thrifty' etc. and adapt the same framework for defining the semantical characteristics, and implementing a Deep Learning model to estimate the likelihood of an image possessing that character.