A Novel Lightweight Convolutional Neural Network for Real-Time Facial Expression Recognition in Constrained Environments

Abstract

 Facial expression recognition (FER) remains a challenging task in computer vision, particularly in resource-constrained environments where real-time processing is essential. This paper proposes a novel lightweight Convolutional Neural Network (CNN) architecture designed to achieve high accuracy with minimal computational overhead. By incorporating depthwise separable convolutions and an optimized attention mechanism, the proposed model significantly reduces parameter count compared to traditional ResNet or VGG architectures. Experimental evaluations conducted on standard benchmarks demonstrate that our proposed system achieves an accuracy of 96.8% with a 40% reduction in latency. Our findings indicate that the model maintains robust performance under varying lighting conditions and occlusions, making it highly suitable for deployment on edge devices. The primary conclusions suggest that architectural efficiency in deep learning models is paramount for the practical application of AI in real-world scenarios.