Repository logo
 

Facial Expression Classification Using R-CNN Based Methods

Files

Whole thesis(2.53 MB)

Date

2019

Authors

Supervisor

Lai, Edmund
Yan, Wei Qi

Item type

Thesis

Degree name

Master of Computer and Information Sciences

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

With the rise of artificial intelligence technology and the development of human vision research, most of researchers are gradually putting more and more attention on the machine recognition of face images. In this thesis, we will study facial expression recognition based on deep learning algorithm. Facial expression is the display of one or more movements or states of facial muscles. Facial expression recognition technology is primarily used in human-computer interaction, intelligent control, security, medical, communication and other fields. In this research, we utilize two deep learning algorithms to implement facial expression recognition system. The first recognition algorithm based on Faster R-CNN, which consists of a fully convolutional network and detector over a region of interest. The second algorithm is Mask R-CNN, which is an extension of Faster R-CNN algorithm that performs image segmentation. Facial expressions are divided into seven categories: anger, contempt, disgust, fear, happy, sadness and surprise. They have been used for object detection and recognition but have not been applied to facial expression classification before. Our experiments show that, compared with the conventional methods, these methods avoided the tedious manual feature extraction, reduced the number of parameters and significantly improved the recognition rate. Moreover, the performance of the trained model in the more realistic settings where the position and angle of the face, lighting, background, etc. are varied are reported in this work.

Description

Keywords

Facial expression recognition, Faster R-CNN, Mask R-CNN, ROI Pooling, ROI Align

Source

DOI

Publisher's version

Rights statement

Permanent link

https://hdl.handle.net/10292/12161

Collections

Masters Theses