This repository provides details of my paper FerNeXt model for facial expression recognition, as presented in the paper "FerNeXt: Facial Expression Recognition Using ConvNeXt with Channel Attention" (IEEE MIUCC Conference) , 2024).
Abstract:
This paper proposes FerNeXt, a facial expression recognition method based on the ConvNeXt architecture, enhanced with an Efficient Channel Attention(ECA) block to improve feature focus. An Affinity loss function is introduced to optimize class separability. The method was evaluated on the AffectNet and RAF-DB datasets, demonstrating improved accuracy and outperforming existing approaches in facial expression recognition.