Hybrid-DS-SCNet: Lightweight module for facial emotion detection

Face Emotion Recognition is a crucial aspect of human-computer interaction, enabling applications in healthcare, security, and affective computing. This paper introduces Hybrid-DS-SCNet, an extremely light-weight Depthwise Separable Convolutional Neural Network with skip connections for detecting and analyzing emotions through low-resolution images. The model integrates DSCNN layers, offering advantages in terms of parameter efficiency and computational cost. With 320K parameters, Hybrid-DS-SCNet shows a great effectiveness on the FER2013 dataset with an impressive accuracy of 74.4%. Its compactness makes it well suited for edge deployment in domains such as mental health monitoring and security.