Facial Expression Recognition Based on Well-Known ConvNet Architectures

Main Article Content

Taima Alrimy
Ahad Alloqmani
Abrar Alotaibi
Nada Aljohani
Salma Kammoun

Abstract

The Convolution Neural Network (CNN) is the most widely used deep learning architecture as it has broken most world records for recognition tasks. Facial Expression Recognition (FER) systems that use classical feature-based techniques, especially CNN’s, is best for classifying images. This paper used three CNN-based methods, which are VGG-16, Inception-v3, and Resnet50-V2 network architectures, to classify facial expressions into seven classes of emotions: happy, angry, neutral, sad, disgust, fear, and surprise. The face expression dataset from Kaggle and JAFFE dataset were used to compare the accuracy between the three architectures to find the pretrained network that best classifies models. The results showed that VGG-16 network architecture produced a higher accuracy (93% in JAFFE and 54% in Kaggle) than the other architectures.


 

Article Details

How to Cite
Alrimy, T., Alloqmani, A. ., Alotaibi, A. ., Aljohani, N. ., & Kammoun, S. . (2022). Facial Expression Recognition Based on Well-Known ConvNet Architectures. Journal of King Abdulaziz University: Computing and Information Technology Sciences, 11(2), 33–43. https://doi.org/10.4197/Comp.11-2.3
Section
Articles