Automate recognition of facial expressions

• Main Author: Zhang, Yuehan
• Other Authors: Lu Shijian
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/147928
• Institution: Nanyang Technological University

Facial expression recognition (FER) has been a prominent research subject for decades. While learning human emotional purpose is a challenging activity that necessitates a huge amount of data, time, and complicated computations. Recently, with the shift of FER from laboratory-controlled to real-life applications and the recent progress of research in deep learning techniques, numerous studies in the automatic facial expression analysis field have been conducted due to its wide application. However, one of the major challenges in FER remains: performance is hampered by uncertainties in large-scale datasets obtained from unconstrained scenarios like FER2013. In this project, Automatic Recognition of Facial Expressions, we explore methods to improve FER system performance on large-scale unconstrained datasets. We firstly introduce existing novel deep neural networks and related training strategies. Then we try to reproduce some state-of-the-art classification results and discuss their advantages and limitations based on datasets and methods applied. After the exploration, we identify some constraints of these methods and try to figure out the cause, base on which, we propose our own ideas to address the constraints.