Hybrid CNN-Transformer Model For Facial Affect Recognition In the ABAW4 Challenge

TL;DR

This paper presents a hybrid CNN-Transformer model designed for facial affect recognition, demonstrating improved performance in the ABAW4 challenge through multi-task learning and synthetic data training.

Contribution

The paper introduces a novel hybrid CNN-Transformer architecture specifically for affect recognition, combining multi-task learning and synthetic data to enhance accuracy.

Findings

Outperforms baseline models on validation data

Verifies effectiveness of the hybrid architecture

Shows benefits of synthetic data in training

Abstract

This paper describes our submission to the fourth Affective Behavior Analysis (ABAW) competition. We proposed a hybrid CNN-Transformer model for the Multi-Task-Learning (MTL) and Learning from Synthetic Data (LSD) task. Experimental results on validation dataset shows that our method achieves better performance than baseline model, which verifies that the effectiveness of proposed network.