Multimediate Challenge

Download

multimediate-challenge@hcics.simtech.uni-stuttgart.de

May 2022

1 participants
2 discussions

Baselines released
by Dominike Thomas 26 May '22

26 May '22

Dear Participants, we apologize for the delay in providing the baseline evaluations. They are now available at https://git.opendfki.de/philipp.mueller/multimediate22_baselines.git These baselines are very simple and we would expect your approaches to outperform them. For your paper submissions, we would like to underline that scientific contribution and novelty is the most important evaluation criterion. I.e. in case your paper provides great insight into the task or tests an interesting method idea this is more valuable than achieving maximum performance with already established approaches. If you currently focus on the latter one, please consider to contribute towards scientific understanding by e.g. conducting interesting ablation experiments or by incorporating a novel idea. All the best for developing your solutions! The MultiMediate’22 Organisers

1 0

Pre-computed Features for the Backchannel Tasks Available
by Dominike Thomas 06 May '22

06 May '22

Hi everyone, We extracted OpenPose[1], OpenFace[2], and OpenSmile[3] features for the backchannel detection and agreement estimation tasks. They are available now on the password-protected dataset download page. The same pre-computed features will be available on the test set as well. We will follow up with features for last year’s tasks. [1] https://github.com/CMU-Perceptual-Computing-Lab/openpose [2] https://github.com/TadasBaltrusaitis/OpenFace [3] https://audeering.github.io/opensmile-python/ Have a nice weekend, Multimediate Organizers

1 0