Here are some datasets we collected for research purpose.
Mobile Spontaneous Affect Response Video (MSARV):
The dataset consists of 11 Asian test subjects. MSARV contains segments of spontaneous facial affects, captured on a mobile device. Elicitation videos are presented to the subject, and the front camera of the mobile device is used to capture the facial response. This produces a response video at 480×640 resolution and 30 frames per second. In total, the dataset contains 817,080 frames.
Video segments used for emotion elicitation include amusing scenes from the comedy “Gags”, talks on popular technology from “Engadget”, academic lectures on advanced topics, sad scenes from “Grey’s Anatomy” and “Les Misérables”, eye and ear surgeries, trailers from horror and ghost movies, and video clips depicting abuse of pregnant woman, children and elderly people.We assembled different segments into two elicitation videos, each approximately 40 minutes and containing 25 short segments. The content of each video is selected to elicit the following affects: happiness (1’30”x3), interest (1’8”x6), boredom (1’15”x5), sadness (2’19”x3), disgust (2’13”x2), fear (2’14”x3) and anger (1’32”x3). (The numbers in the brackets indicate the average length and the number of segments.) These are the same affects that are covered in most of the publicly-available datasets. Some of the affects (e.g. happiness) are more easily aroused than others (e.g. sadness) , which accounts for the difference in the length and number of the elicitation videos. Segments are kept between 1-2 min to avoid habituation to the stimuli while being long enough to arouse an affect .
The paper introducing the dataset:
Michael Xuelin Huang, Grace Ngai, Kien A. Hua, Stephen C.F. Chan and Hong Va Leong,”Identifying User-Specific Facial Affects from Spontaneous Expressions with Minimal Annotation”, IEEE Transactions on Affective Computing.
To acquire the dataset, please print the End User License Agreement (EULA, download below), sign it, scan it and send it back to firstname.lastname@example.org.
Please note, we only accept requests from academic (i.e. university) email-addresses. Any requests from free email addresses (hotmail, yahoo, gmail etc) will be refused.