Frequently Asked Questions

Challenge registration

  • We completed the registration form, what are the next steps?

Once your form is validated, you will receive an email with the EULA(s) corresponding to the Sub-challenge(s) you registered.

  • We previously completed and send an EULA for the DAIC/SEWA database, shall we send another EULA for AVEC 2019?

Yes please. The datasets have been updated. You might also want to update the names of the team members appearing on the EULA.

  • Is there any restriction on the number of team members?

There is no restriction on the number of participants in a team.

  • Is it possible to appear in several teams?

It is possible to appear in several teams, even as a same team leader. However, the number of five trials for submissions on the test sets will be shared amongst all the teams.

System development

  • Can we use others datasets or pre-trained models?

The use of external resources is allowed for all Sub-challenges. Please include a description of the used additional datasets/models when writing your paper.

  • Since the AVEC 2019 DDS is an extension of the AVEC 2016 DSC, has the 2016 data already been included?

The DAIC-WOZ (train/dev/test), which was indeed used in AVEC 2016, is constituting the training and development sets of the AVEC 2019 DDS. This year development partition also includes subjects from the newly collected data.

  • When the transcriptions will be released?

We do not provide speech transcripts to make sure text data are processed in the same conditions as the others modalities, i.e., without any human intervention that facilitates their analysis. Please therefore use ASR from the audio signal if you want to use textual information.

  • For the SoMS, will you evaluate the initial mood predictions, i.e. values given as metadata and corresponding to the self-report made before the narrative stories? Will those initial ratings be provided in the test set?

We do not evaluate the prediction of the initial self-report of mood, only the self-reports that were made after the narrative of the personal stories. As the main interest of the SoMS is on the change in mood, rather it's static analysis, initial self-reports made before the narrative stories of the subjects included in the test set are provided and can be thus exploited as contextual information.

  • For CES of AVEC2019, because our team want to use semi-supervised learning method, I am writing to inquire if we can use the test set data as unlabeled data to train our model.

You can indeed use the test set data to perform any unsupervised or semi-supervised processing on it, as long as you do not manually label those files.

Test results submission

  • How many times can I submit my results?

Each team can submit results five times per Sub-challenge; badly formatted results are not counted towards one of your five submissions.

  • I’ve submitted a set of results, when will I receive my scores?

During an active challenge we strive to return scores within 24hrs during typical working days (Monday – Friday), however please try to be patient as this is subject to workload. Under no circumstances should you spam the organisers as this simply delays the process for all teams.

  • I have submitted all my results on the test partition, can I know my rank now?

Results of the AVEC 2019 Challenge will be announced at the end of the Workshop's day.

  • Will we be ranked with the mean of our best attempt, or will you first select our best scores independently of the submission?

For the CES, where several emotional dimensions are tested, the ranking is based on the mean of the best CCC score obtained on the emotional dimensions independently of the attempt. There is therefore no need to submit a trial with the best systems tuned for each dimension.

  • Are test instances of the German and Hungarian cultures, in addition to those of the Chinese culture, used for ranking?

Only test instances of the Chinese culture are used for ranking; the CES is about cross-cultural inference with no information on the tested culture, which is only the case for the Chinese culture since labels are provided for the German and Hungarian cultures.

Paper submission

  • We registered a team and would like to submit our work on the AVEC 2019 datasets to another workshop/conference, is this possible?

You are free to submit your results to any venue. However, only teams that submitted a paper to the AVEC 2019 Workshop, and at least one trial on the test set, will be eligible for ranking in the Challenge.