HSEmotion Team at the 6th ABAW Competition: Facial Expressions, Valence-Arousal and Emotion Intensity Prediction

TL;DR

This paper explores using pre-trained lightweight deep models to improve facial emotion analysis in-the-wild, achieving significant performance gains without extensive fine-tuning across multiple affective recognition tasks.

Contribution

The study introduces multi-task trained lightweight models based on MobileViT, MobileFaceNet, EfficientNet, and DDAMFN for facial expression and emotion recognition, enhancing accuracy without fine-tuning.

Findings

Significant improvement in validation metrics over existing methods.

Effective multi-task training of lightweight models for emotion recognition.

Successful application across five affective analysis tasks.

Abstract

This article presents our results for the sixth Affective Behavior Analysis in-the-wild (ABAW) competition. To improve the trustworthiness of facial analysis, we study the possibility of using pre-trained deep models that extract reliable emotional features without the need to fine-tune the neural networks for a downstream task. In particular, we introduce several lightweight models based on MobileViT, MobileFaceNet, EfficientNet, and DDAMFN architectures trained in multi-task scenarios to recognize facial expressions, valence, and arousal on static photos. These neural networks extract frame-level features fed into a simple classifier, e.g., linear feed-forward neural network, to predict emotion intensity, compound expressions, action units, facial expressions, and valence/arousal. Experimental results for five tasks from the sixth ABAW challenge demonstrate that our approach lets us significantly improve quality metrics on validation sets compared to existing non-ensemble techniques.