Dynamics of Collective Group Affect: Group-level Annotations and the Multimodal Modeling of Convergence and Divergence

This repository contains the codebase for the paper:

Navin Raj Prabhu, Maria Tsfasman, Catharine Oertel, Timo Gerkmann, Nale Lehmann-Willenbrock, "Dynamics of Collective Group Affect: Group-level Annotations and the Multimodal Modeling of Convergence and Divergence", Under Review, Submitted to IEEE Transactions on Affective Computing, Sep, 2024. arxiv

The repository contains the following:

Data and analysis

1. Data Reader and Classes
2. Annotation Analysis
3. Inter-annotator agreement
4. Ground-truth consensus

Modeling

1. Individual-level feature extractors
   1. Audio (prosody)
   2. Video
2. Dyad-level feature extractors (Jobs for feature extraction available)
   1. Synchrony (correlation, lagged correlation)
   2. Convergence (global, symmetric and asymmetric)
3. Convergence and Divergence analysis
4. Group Affect recognition baselines (Use configs for baseline versions)