Important information for your evaluations!
This message contains several important bits of information for your evaluations. Please read it carefully.
(1) If you experience difficulties with obtaining results on eval.ai, please send your test set predictions directly to Michael Dietz and Dominik Schiller. They will evaluate the predictions for you.
(2) We found a bug in our baseline evaluation code. As a result, there are new baseline results for the agreement estimation subchallenge (please see https://git.opendfki.de/philipp.mueller/multimediate22_baselines). We apologize if this causes any inconvenience on your side!
(3) Please report results on the official challenge test set in your paper submission. For ablation studies please make use of the official validation set to ensure comparability to other approaches as well as the official baselines.
(4) We added a few more ablation results for the baselines (see https://git.opendfki.de/philipp.mueller/multimediate22_baselines). Please cite the following publication in your paper to refer to the official baselines and the challenge in general:
Philipp Müller, Michael Dietz, Dominik Schiller, Dominike Thomas, Hali Lindsay, Patrick Gebhard, Elisabeth André, and Andreas Bulling. 2022. MultiMediate ‘22: Backchannel Detection and Agreement Estimation in Group Interactions.