Detecting expressions with multimodal transformers

TL;DR

This paper explores deep learning models, including transformers, for detecting person expressions through audio-visual cues, demonstrating improved performance over recurrent models and emphasizing the importance of visual data.

Contribution

It introduces a transformer-based architecture for audio-visual expression detection, outperforming recurrent models and highlighting multimodal integration benefits.

Findings

Transformer models outperform recurrent baselines by ~2% in arousal and valence detection.

Multimodal models outperform single modality models by up to 3.6%.

Visual modality is crucial for accurate expression detection.

Abstract

Developing machine learning algorithms to understand person-to-person engagement can result in natural user experiences for communal devices such as Amazon Alexa. Among other cues such as voice activity and gaze, a person's audio-visual expression that includes tone of the voice and facial expression serves as an implicit signal of engagement between parties in a dialog. This study investigates deep-learning algorithms for audio-visual detection of user's expression. We first implement an audio-visual baseline model with recurrent layers that shows competitive results compared to current state of the art. Next, we propose the transformer architecture with encoder layers that better integrate audio-visual features for expressions tracking. Performance on the Aff-Wild2 database shows that the proposed methods perform better than baseline architecture with recurrent layers with absolute gains approximately 2% for arousal and valence descriptors. Further, multimodal architectures show significant improvements over models trained on single modalities with gains of up to 3.6%. Ablation studies show the significance of the visual modality for the expression detection on the Aff-Wild2 database.