Utilizing deep transfer learning models and data augmentation to improve image classification

Authors

  • Inderpreet Kaur, Amanpreet Kaur Sandhu, Yogesh Kumar

DOI:

https://doi.org/10.17762/msea.v71i3.1515

Abstract

VBDs (vector-borne diseases) have a significant influence on people's health and economies all around the globe. Malaria is a disease spread by female Anopheles mosquitoes, which give the host a motile infectious version of the disease. Malaria is characterised by fever, headache, fatigue, and vomiting. In extreme circumstances, it might cause coma and death. The most of these deaths are caused by a delayed or incorrect diagnosis. Malaria is best diagnosed using a manual microscope. However, it is labour-intensive and prone to errors caused by humans. This is a severe worldwide health risk; hence it is essential that the review process be automated. Therefore, a highly accurate automated computational approach is needed to aid in the timely identification of malaria in order to lower fatality rates. In order to improve diagnosis accuracy, researchers are using deep-learning technologies like convolutional neural network (CNNs) and image processing to assess Parasitemia in tiny blood slides. ResNet50, VGG-16, and VGG-19 are three CNN architectures that were trained using the same dataset and are fed images of both infected and uninfected erythrocytes. Fine-tuning is contrasted with transfer learning in terms of its consequences. Based on the evaluated parameters and data set, the VGG-19 model had the highest accuracy.

Downloads

Published

2023-01-09

How to Cite

Inderpreet Kaur, Amanpreet Kaur Sandhu, Yogesh Kumar. (2023). Utilizing deep transfer learning models and data augmentation to improve image classification. Mathematical Statistician and Engineering Applications, 71(3), 1923–1932. https://doi.org/10.17762/msea.v71i3.1515

Issue

Section

Articles