Deep Learning Framework for Facial Emotion Recognition using CNN Architectures

Abstract

FER (facial expression recognition) is a significant study subject in the artificial intelligence and computer vision areas because of its widespread applicability in both academic and industrial sectors. Though FER can be carried out primarily utilizing multiple sensors, research shows that using facial images/videos for recognition of facial expression is better because visual expressions carry major information through which emotions can be conveyed.In the past, much research was conducted in the field of FER using different approaches such as the use of different sensors, machine learning, and deep learning framework with dynamic sequences and static images. The most recent state-of-the-art outcomes demonstrate In comparison to conventional FER techniques, deep learning Convolutional Neural Network (CNN) based systems are significantly more powerful. Deep learning-based FER methods utilizing deep networks enable extraction of features automatically instead of traditionally handcrafted feature extraction.This paper focuses on implementing different custom and standard CNN architectures for training and testing them on facial expression static image datasets scenario KDEF, RAFD, RAF-DB, SFEW, and AMFED+, both lab-controlled and wild.