2017 Looking at People ICCV Challenge - Fake vs. true facial emotion recognition

Track description

Track Information

The database contains video set of 50 different subjects. There are 12 videos for each subject. Each video represents one of basic emotions (Angry, Contempt, Disgust, Happy, Sad, Surprised), for both the true and fake emotions. A sample sequence is shown in Fig. 1. The database has been divided to three mutually exclusive subsets:

Each set includes in average 12 videos with transition either from neutral condition to the true emotion or from true emotion to a fake emotion. Subjects were made to watch videos which induced the emotions. After that the subjects were asked to act the emotions felt accordingly.

Fig. 1. A sample video sequence for fake and true emotions

Main Task:

1. Recognition of fakeness of an emotion;
2. Recognition of fake/true specific emotion

Participants can register and participate in this track through https://competitions.codalab.org/competitions/16611.