BlEmoRe Competition on Multimodal Blended Emotion Recognition Call for Papers

BlEmoRe Competition on Multimodal Blended Emotion Recognition at IEEE FG'26
https://blemore.tech/

Humans often experience multiple emotions simultaneously, such as
feeling both sadness and anger when facing an unjust loss or happiness
and surprise at a birthday party. However, emotion recognition
research has largely focused on single emotions due to limited
datasets and a lack of awareness. The BlEmoRe competition aims to
bridge this gap by advancing the recognition of multimodal blended
emotional expressions. BlEmoRe introduces a novel dataset of
multimodal emotion expressions that contains both single emotions and
blended emotions conveyed with similar and varying proportions. We
invite submissions addressing the challenging task of blended emotion
recognition on this novel dataset.

How to Participate

Training data and unlabelled test data is already available
(https://zenodo.org/records/17787362). Participants submit predictions
on the test data for evaluation on our servers. We employ two
evaluation metrics:

    ACCpresence measures whether the correct label(s) are predicted
    without errors. A correct prediction must include all present
    emotions while avoiding false negatives (e.g., predicting only one
    emotion in a blend of two emotions) and false positives (e.g.,
    predicting emotions that are not part of the label).

    ACCsalience extends ACCpresence by considering the relative
    prominence of each emotion. It evaluates whether the predicted
    proportions reflect the correct ranking. This metric applies only
    to blended emotions.

Baseline approaches and evaluation code are available on GitHub
(https://github.com/BlEmoRe/blemore-common).

Tentative Dates

    Data available: Already available!

    Results submission: Early March 2026

    Paper submission: Mid March 2026

    Conference: 25-29 May 2026

Organisers

Tim Lachmann (Stockholm University)

Philipp Müller (Max Planck Institute for Intelligent Systems)

Teimuraz Saghinadze (Georgian Technical University)

Michal Balazia (INRIA Université Côte d’Azur)

Alexandra Israelsson (Stockholm University)

Petri Laukka (Uppsala University)