Comparative Analysis of Deep Convolutional Neural Networks Architecture in Facial Expression Recognition: A Survey

Facial expression recognition (FER) has made much progress and is supported by many scientific studies conducted in the past decade. The technique and architecture model used in the FER are the aspects that get the most improvement from the researchers. One of the famously used techniques in conducting FER is Deep Convolutional Neural Network (DCNN). The development of DCNN architecture has a vital role in increasing the accuracy of facial expression recognition. The choice of architecture will also affect the total computational costs required to perform facial expression recognition activities. This paper compares some of the DCNN architectures in the past FER research during 2010-2020 using various FER dataset. This paper presents detailed information on several DCNN architectures in terms of the dataset preprocessing techniques and accuracy value for some accessible FER datasets such as Fer2013, CK+, AFEW, and other datasets. This research will explain other FER researchers who are new in FER research to determine the DCNN architecture used based on several FER datasets.