Multimediate Challenge

Download

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

May 2021

1 participants
1 discussions

Full annotations for next speaker prediction and eye contact annotations fix
by Dominike Thomas 21 May '21

21 May '21

Good evening, The full fine-grained speaking annotations are available now! If you did not obtain the dataset yet, please visit https://multimediate-challenge.org/Dataset/ for instructions. In case you already received download links to the dataset, you can simply re-download the annotations.zip file. IMPORTANT: There is an additional change in the eye contact annotations because of correction of a single annotation error (sample 942 on the training set). Here it was the case that the person at subject position 3 was labelled to be looking at herself. We corrected the annotation (the actual gaze target is position 4). We encourage you to also update your "train_eye_contact.csv" by downloading the corrected annotation files. Please do not hesitate to contact us if you have any questions. Dominike Thomas Universität Stuttgart | University of Stuttgart Abteilung Mensch-Computer-Interaktion und Kognitive Systeme | Department Human-Computer Interaction and Cognitive Systems (HCICS) Institut für Visualisierung und Interaktive Systeme (VIS) | Institute for Visualization and Interactive Systems (VIS) Pfaffenwaldring 5a, 01.027 70569 Stuttgart Deutschland | Germany +49 (0) 711 685-60051 dominike.thomas(a)vis.uni-stuttgart.de<mailto:dominike.thomas@vis.uni-stuttgart.de> https://www.perceptualui.org/

1 0