Abstract

The Audio/Visual Emotion Challenge and Workshop (AVEC 2018) “Bipolar Disorder and Cross-cultural Affect” is a satellite event of ACM MM 2018, (Seoul, Korea, 22 October 2018), and the eighth competition aimed at comparison of multimedia processing and machine learning methods for automatic audio, visual, and audio-visual health and emotion sensing, with all participants competing under strictly the same conditions.

The goal of the Challenge is to provide a common benchmark test set for multimodal information processing and to bring together the audio, visual and audio-visual affect recognition communities, to compare the relative merits of the approaches to automatic health and emotion analysis under well-defined conditions. Another motivation is the need to advance health and emotion recognition systems to be able to deal with fully naturalistic behaviour in large volumes of un-segmented, non-prototypical and non-preselected data, as this is exactly the type of data that both multimedia and human-machine/human-robot communication interfaces have to face in the real world.

The challenge will showcase the best results obtained by the teams who participated in the following Sub-Challenges:

Bipolar Disorder Sub-Challenge

Patients suffering from bipolar disorder – as defined by the DSM-5 – needed to be classified into remission, hypo-mania, and mania, from audio-visual recordings of structured interviews (BD corpus); performance is measured by the unweighted average recall (UAR) over the three classes.

Cross-cultural Emotion Sub-Challenge

Dimensions of emotion needed to be predicted time-continuously in a cross-cultural setup (German => Hungarian) from audio-visual data of dyadic interactions (SEWA corpus); performance is the concordance correlation coefficient (CCC) averaged over the dimensions.

Gold-standard Emotion Sub-Challenge

Individual annotations of dimensional emotions needed to be processed to create a single time series of emotion labels termed as “gold-standard”. Performance (CCC) is measured by a baseline system trained and evaluated from multimodal data with the generated gold-standard (RECOLA corpus), and a condition on the unexplained variance between this time series and the (original) individual annotations was used for validation of the results.

Read more about the challenge guidelines.

Read more about the Challenge's scope and topics in this section.

Download the paper introducing the challenge.

Read more about the program.