Paper Title
A New Deep Learning Approach with a Hybrid of Light Emotion Network and Ensemble Model for Emotion Recognition from Facial Expressions

Abstract
In this study, a new deep learning model that combines light computational cost and high accuracy for emotion recognition from facial expressions is proposed. The proposed model is based on Light Emotion Network and Hybrid Ensemble architecture, effectively combining CBAM attention mechanism and MobileViT blocks. Experiments conducted on the Facial Expression Recognition dataset obtained from Kaggle have shown that the model achieves an average accuracy of 95.1% on seven basic emotion classes. The distinguishing feature of the model is that it can provide high performance with limited computational resources. The model can produce consistent results despite illumination conditions, pose variations, and partial occlusion cases. The results of this study provide a promising basis for real-time emotion analysis applications.