Date of Award
Dr. Dimitrios Charalampidis
Dr. Abdul Rahman Alsamman
Dr. Kim Jovanovich
Facial emotion recognition is a widely studied area with applications in diverse domains such as human-computer interaction, affective computing, and social robotics. This thesis aims to improve the accuracy of facial emotion recognition models by incorporating a second neural network trained on original probabilities and probability transformation, while also comparing the performance of different techniques. The thesis begins with a thorough review of available datasets and technologies used for data collection, highlighting the challenges associated with these datasets. A detailed analysis of various facial emotion detection models, including the baseline model and its different architectures, is presented. The thesis also explores the pre-processing of datasets for binary classifiers and investigates the effects of developing an ensemble of binary classifiers.The main contribution of the thesis is the incorporation of a second neural network trained on the probabilities of binary models, along with probability transformation, to enhance the accuracy of facial emotion recognition models. Experimental results on the FER2013 dataset are presented, demonstrating the effectiveness of this approach, achieving a best accuracy of 69.4%. Additionally, the thesis compares the performance of different techniques to provide insights into their relative effectiveness in improving facial emotion recognition accuracy.The thesis concludes with a summary of the results, drawing conclusions from the analysis, and discussing future directions for further research in facial emotion recognition. The findings of this research contribute to the advancement of facial emotion recognition techniques and provide valuable insights for researchers and practitioners in the field.
Hanin, Arsany, "Comparison of Facial Emotion Recognition Models Using Deep Learning" (2023). University of New Orleans Theses and Dissertations. 3076.