No abstract

Creating large, richly annotated databases depicting real-world or simulated real-world conditions is a challenging task. There has been a long understood need for recognition of human facial expressions in realistic video scenarios. Although many expression databases are available, research has been restrained by their limited scope due to their 'lab controlled' recording environment. This paper proposes a new temporal facial expression database Acted Facial Expressions in the Wild (AFEW) and its static subset Static Facial Expressions in the Wild (SFEW), extracted from movies. As creating databases is time consuming and complex, a novel semi-automatic approach via a recommender system based on subtitles is proposed. Further, experimental protocols based on varying levels of person dependency are defined.AFEW is compared with the extended Cohn-Kanade CK+ database and SFEW with JAFFE and Multi-PIE databases Index TermsFacial expression recognition, large scale database, real-world conditions, emotion database

Quality data recorded in varied realistic environments is vital for effective human face related research. Currently available datasets for human facial expression analysis have been generated in highly controlled lab environments. We present a new static facial expression database Static Facial Expressions in the Wild (SFEW) extracted from a temporal facial expressions database Acted Facial Expressions in the Wild (AFEW) [9], which we have extracted from movies. In the past, many robust methods have been reported in the literature. However, these methods have been experimented on different databases or using different protocols within the same databases. The lack of a standard protocol makes it difficult to compare systems and acts as a hindrance in the progress of the field. Therefore, we propose a person independent training and testing protocol for expression recognition as part of the BEFIT workshop. Further, we compare our dataset with the JAFFE and Multi-PIE datasets and provide baseline results.

No abstract