Evaluations on the challenge test sets are now possible also for this years tasks (bodily behaviour recognition and engagement estimation). Test set evaluations on MultiMediate’21 and ’22 tasks were already, and will continue to be possible by sending your predictions to Michael Dietz. In the following we provide information on how to evaluate your models for this years’ novel tasks.
Evaluation is performed by submitting predictions to Eval.ai. Please do the following: