Call for contributions
*****IUI Dyadic IMPRESSION Recognition Challenge (virtual event)*****
https://simsimpression.unige.ch/
IMPORTANT DATES
26/10/2021: Start of the Challenge, release of training data
07/01/2022: Abstract submission (validation results) and release of the test data
24/01/2022: Final paper submission - End of the competition (test results)
09/02/2022: Notification of paper acceptance
22/03/2022: Workshop held (online)
MOTIVATION
The Dyadic IMPRESSION Recognition Challenge, to be held in March 2022 in conjunction with IUI 2022 in Helsinki, Finland, will be devoted to all aspects of artificial intelligence and behavorial science for the analysis of human-human interaction from multimodal data.
To advance and motivate the research on human bodily responses in dyadic interactions, we organize the challenge which uses the open and accessible multimodal IMPRESSION dataset. It addresses multimodal recognition as well as dynamic multi-user recognition, where both interlocutors’ information can be exploited.
THE CHALLENGE
The challenge aims at automatic impression recognition. This challenge will focus on automatic impression recognition of multiple individuals (i.e., the Receiver, that is, the person who forms an impression of the other, i.e., the Emitter) during a dyadic interaction. Self-reported impressions in the warmth and competence dimensions are given, associated with synchronized face videos, eye movements and physiological signals (including ECG, BVP and GSR) of both Emitters and Receivers.
The challenge is composed of two phases:
Development phase: public training data will be released and participants will develop their approaches and validate their predictions using a validation set;
Test (final) phase: participants will need to submit their predicted targets with respect to the test data, which will be released just a few days before the end of the challenge. We will then rank submissions by performance and communicate the results during the workshop.
The evaluation consists of computing the average concordance correlation coefficient (CCC) among the participants tested for the warmth and competence between the predicted continuous values and the continuous ground truth values.
All participants are invited to submit papers describing their solution to the challenge and following the workshop submission guidelines of the IUI conference.
Examples of potential submissions include (but are not limited to):
the combination of multimodal measures from either the Receiver or Emitter;
the computation of synchrony features between the Receiver and Emitter;
deep learning architectures which combine features from the Receiver and Emitter;
transfert learning approaches to extract features;
comparative studies of several approaches.
THE DATASET
The IMPRESSION dataset aims to focus on the development of automatic approaches to study and understand the mechanisms of perception and adaptation to verbal and nonverbal social signals in dyadic interactions, taking into account individual and dyad characteristics. To the best of our knowledge, there is no similar publicly available, nonacted face-to-face dyadic dataset in the research field in terms of participants, recorded sessions, and continuous impression labels in warmth and competence. Detailed information about the IMPRESSION dataset is provided on the challenge website and in the following paper:
https://archive-ouverte.unige.ch/unige:155675
Challenge Contact email:
Organizers:
Chen Wang, University of Geneva, Switzerland
Guillaume Chanel, University of Geneva, Switzerland
Beatrice Biancardi, LTCI, Télécom Paris, France
Chloé Clavel, LTCI, Télécom Paris, France