Sigatoka and Xanthomonas Banana Leaf Disease Detection Via Transfer Learning

Articles in Press

Document Type : Article

Authors

1 Department of Software Engineering, CoE for HPC and BDA, AASTU, Addis Ababa, Ethiopia

2 Department of Software Engineering, CoE for HPC and BDA, AASTU, Addis Ababa, Ethiopia,

3 School of Information Technology and Engineering, AAiT, Addis Ababa, Ethiopia

4 Faculty of Emerging Technologies, Sri Sri University, Cuttack, Odisha, India

Abstract

Plant diseases are one of the leading causes of famine
and food insecurity worldwide. Furthermore, the diseases are
thought to be responsible for up to 16% of annual agricultural
output losses globally. Bananas are a widely grown crop and
a popular fruit in both developed and developing countries.
However, diseases like Xanthomonas Wilt and Sigatoka leaf spot
are the most serious problems in banana cultivation, with yield
losses of up to 100%. In this study, the CNN algorithm was
used to develop a disease detection model for Xanthomonas Wilt
and Sigatoka Leaf spots in bananas. To do so, a banana plant
leaf image dataset was gathered from the Southern Nation Nationalities
and Peoples Regional State Arbaminch Zuria Woreda
Lante kebele and Chano kebele, as well as the Gamugofa zone
Mierab Abaya Woreda Omolante kebele, where bananas are
widely grown and Xanthomonas Wilt and Sigatoka leaf spot
diseases, are highly observed. To solve the limitation of the
dataset the study applied data augmentation techniques. Furthermore,
transfer learning techniques, the MobileNet, EfficientNet,
VGG16, VGG19, and InceptionV3 were employed. While testing
with test data, the VGG16 model scored 81.53% which is higher
than other pretrained models.

Keywords

Main Subjects

Scientia Iranica

Articles in Press, Accepted Manuscript
Available Online from 12 May 2024

Files

History

  • Receive Date: 25 April 2023
  • Revise Date: 12 January 2024
  • Accept Date: 12 May 2024

Share

How to cite

Statistics