Award Date
August 2023
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Computer Science
First Committee Member
Fatma Nasoz
Second Committee Member
Kazem Taghva
Third Committee Member
Jan Pedersen
Fourth Committee Member
Mingon Kang
Fifth Committee Member
Emma Regentova
Number of Pages
218
Abstract
Facial expressions play a crucial role in human communication, serving as a powerful means to convey emotions. However, classifying facial expressions using artificial intelligence (AI) can be challenging, especially with small datasets and images. Facial Expression Recognition (FER) is an active area of research, with Convolutional Neural Networks (CNNs) being widely employed for classification. In this research, we propose a CNN-based approach for FER that utilizes both original and augmented datasets to enhance classification accuracy. Experimental results on the FER2013 dataset show test accuracies of 63.39% and 64.59% for the original and augmented datasets, respectively, in a seven-class classification task. These results were compared to the state-of-the-art classification, which achieved a 75.2% accuracy. Our approach incorporates data augmentation techniques to mitigate overfitting and improve the performance of the CNN classifier. Additionally, we aim to bridge the gap between supervised and unsupervised learning in FER by integrating Generative Adversarial Networks (GANs) into our approach. GANs offer a unique approach to generate synthetic images and have been widely used in various computer vision applications. We provide a theoretical basis for GANs and demonstrate their application in generating realistic synthetic images using the FER2013 dataset. Furthermore, we replicate three popular GAN architectures using the MNIST dataset, validating the feasibility of using GANs for image generation and data augmentation in the context of FER tasks. Overall, our proposed approach combines CNNs for classification with GANs for data augmentation, aiming to improve the accuracy of FER models while mitigating overfitting. This has the potential to advance the understanding of human emotions through AI technologies and contribute to the field of FER. The utilization of GANs in FER research also opens possibilities for generating realistic images from other facial expression datasets, expanding the application of AI in this domain.
Keywords
Data Augmentation; Emotion Detection with Artificial Intelligence-based Expression Analysis; Human-Computer Interaction; Image Classification; Machine / Deep Learning Techniques; Synthetic Image Generation
Disciplines
Artificial Intelligence and Robotics | Computer Engineering | Computer Sciences
File Format
File Size
36050 KB
Degree Grantor
University of Nevada, Las Vegas
Language
English
Repository Citation
Singh, Shekhar, "Facial Expression Recognition Using Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs) for Data Augmentation and Image Generation" (2023). UNLV Theses, Dissertations, Professional Papers, and Capstones. 4852.
http://dx.doi.org/10.34917/36948203
Rights
IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/