SEMAINE was a three year EU FP7 project that ended in December 2010.

The goal of the SEMAINE project was to develop an artificial conversiational partner that understands social signals and responds accordingly, a so-called Sensitive Attentive Listener. Within the project, I work on the computer vision analysis techniques, such as face detection, facial point detection and tracking, and Action Unit detection. The official SEMAINE website contains an official description of the project, as well as all the EU FP7 reports.

The SEMAINE project's two main outputs are a database and a conversational agent. The conversational agent is a so-called SAL agent, for Sensitive Attentive Listener. This means that the agent is able to detect social signals such as head actions, facial expressions, and vocal intonation and uses this information to understand the cognitive state of the dyadic partner. It also creates its own facial expressions, and vocal intonations, and uses this for back-channeling and conveying concepts such as underdstanding and agreement. The full SEMAINE system is available from sourceforge, and installation instructions can be found on the SEMAINE wiki. The video analysis components, created by me and my colleague Hatice Gunes can be downloaded separately from the iBUG website.

The second main output is the SEMAINE database: a large database of people interactinng with SAL characters that are either represented by humans or by different versions of the SEMAINE system.

A number of recorded interactions between users and the final version of SEMAINE are available on youtube: