DAISEE: Dataset for Affective States in E-Learning Environments.

Extracting and understanding a ective states of subjectsthrough analysis of face videos is of high consequence to advance thelevels of interaction in human-computer interfaces. This paper aims tohighlight vision-related tasks focused on understanding reactions ofsubjects to presented content which has not been largely studied by thevision community in comparison to other emotions. To facilitate futurestudy in this eld, we present an e ort in collecting DAiSEE, a free touse large-scale dataset using crowd annotation, that not only simulatesa real world setting for e-learning environments, but also captures theinterpretability issues of such a ective states by human annotators. Inaddition to the dataset, we present benchmark results based on stan-dard baseline methods and vote aggregation strategies, thus providing aspringboard for further research.