White Blood Cell Classification: A Comparison between VGG16 and ResNet50 Model

Abstract

White blood cell classification plays a significant role in helping a physician to diagnose disease. Using automated analyser machine can be easily analyse, fast, and accurate but the machine is very costly. Alternatively, this task can be manually perform by human who are expert in the field. However, it is very laborious. Machine learning and computer vision are applied to solve these limitations. In this study, two Convolutional Neural Networks–VGG-16 and ResNet-50–are employed to classify five types of white blood cell: Basophil, Eosinophil, Neutrophil, Lymphocyte, Monocyte. The results show that ResNet-50 is the best and can achieve 88.29 % accuracy.