We have made over 300 manually coded frame-by-frame annotations of the temporal segments available. The annotations are of sequences in Part I and Part II of the database (see Valstar & Pantic, LREC 2010 for definition of the parts). The annotations are EMFACS annotations. Unfortunately you cannot search for the existence of these annotations at this point. Look for sessions with two EMFACS annotations, one of them will be the onset-apex-offset labelling.

The MMI Facial expression database is now also active on Twitter! Follow @mmidb for updates on this database, including data and annotation additions.

The paper "Induced Disgust, Happiness and Surprise: an Addition to the MMI Facial Expression Database" by Michel F. Valstar and Maja Pantic has been accepted for the workshop on Emotional Corpora, as part of LREC 2010. Soon the EULA of the MMI Facial Expression Database will be changed to state that anyone using data from parts II - V of the database will have to cite this paper.

Part V of the database has been added. This consists of Sessions 2895-2903. Each session is a recording of induced laughter (subject watched comedy clips), and lasts about 20 mintues. The recording scenario type of these Sessions is 'reacted'.

There are 2894 sessions in the database. Currently 1395 sessions are AU coded and 197 sessions are labelled as one of the six basic emotions.