Improving face recognition with ensemble learning: a comparative study of EigenFace, FisherFace, and ResNet-18
Mei Qian, Wu Shixiao · 2025
This Face recognition is a critical task in security and authentication systems. Traditional methods like EigenFace and FisherFace, while effective, struggle with variations in pose and illumination. Deep learning methods, particularly ResNet-18, have shown promise in addressing these challenges but often require large datasets and computational resources. This paper presents an ensemble learning approach that combines EigenFace, FisherFace, and ResNet-18 to enhance recognition accuracy. Experimental results demonstrate that while EigenFace achieves 74.78% accuracy and FisherFace 81.53%, ResNet-18 alone performs at 33.94%. However, by integrating ResNet-18 with EigenFace and FisherFace using ensemble learning, the accuracy improves to 83.53%. These findings highlight the potential of combining classical and deep learning methods to improve face recognition performance.