Comparative Review on Methods of Facial Emotion Recognition

Abstract

Emotion detection using facial recognition is widely used in robotics, psychology, gaming, and security. Differ[1]ent models and methods yield different accuracy and performance depending on the dataset used, their construc[1]tion and the hyperparameters set. This paper includes a comparative based review of two models namely Deep[1]Emotion and MicroExpNet based on Attentional Convolutional Network (ACN) and Knowledge Distillation (KD) respectively which are both based on Convolutional Neural Network (CNN). The paper presents the com[1]parison of the two Facial Emotion Recognition models through a literature review. The comparative review is based on their network architecture, their results, and hyperparameters. This review provides the idea of the applicability of the models in different fields as both of the models proves to be better in their domain. The MicroExpNet model due to its size and speed is known to have a future scope in mobile deployment and provides valuable insights to the development of microarchitecture whereas Deep-Emotion using ACN detects emotion in relatively fewer layers of CNN. This paper further contributes to laying a foundation for future implementers to better understand which model to implement concerning the area of their application.