Exploiting multi-CNN features in CNN-RNN based Dimensional Emotion Recognition on the OMG in-the-wild Dataset

TL;DR

This paper introduces a CNN-RNN framework that leverages multi-level CNN features for dimensional emotion recognition in-the-wild, achieving state-of-the-art results using only visual data on the OMG-Emotion dataset.

Contribution

It proposes a novel multi-level feature extraction and fusion approach within CNN-RNN models, enhancing emotion recognition performance in challenging real-world scenarios.

Findings

Outperformed state-of-the-art methods using only visual data.

Achieved second place in OMG-Emotion Challenge for valence estimation.

Combining low- and high-level features significantly improves arousal estimation.

Abstract

This paper presents a novel CNN-RNN based approach, which exploits multiple CNN features for dimensional emotion recognition in-the-wild, utilizing the One-Minute Gradual-Emotion (OMG-Emotion) dataset. Our approach includes first pre-training with the relevant and large in size, Aff-Wild and Aff-Wild2 emotion databases. Low-, mid- and high-level features are extracted from the trained CNN component and are exploited by RNN subnets in a multi-task framework. Their outputs constitute an intermediate level prediction; final estimates are obtained as the mean or median values of these predictions. Fusion of the networks is also examined for boosting the obtained performance, at Decision-, or at Model-level; in the latter case a RNN was used for the fusion. Our approach, although using only the visual modality, outperformed state-of-the-art methods that utilized audio and visual modalities. Some of our developments have been submitted to the OMG-Emotion Challenge, ranking second among the technologies which used only visual information for valence estimation; ranking third overall. Through extensive experimentation, we further show that arousal estimation is greatly improved when low-level features are combined with high-level ones.