Deep learning approach for Cassava Phytoplasma Disease image classification

Authors

  • Aleth Julianne O. Caringal ⋅ PH Institute of Mathematical Sciences and Physics, University of the Philippines Los Baños
  • Marisol P. Martinez ⋅ PH Institute of Mathematical Sciences and Physics, University of the Philippines Los Baños
  • Sarah Jane B. Manaday ⋅ PH National Crop Protection Center, University of the Philippines Los Baños
  • Malec Renz Jeremiah B. Martirez ⋅ PH National Crop Protection Center, University of the Philippines Los Baños
  • Gideon Aries S. Burgonio ⋅ PH National Crop Protection Center, University of the Philippines Los Baños
  • Melvin D. Ebuenga ⋅ PH National Crop Protection Center, University of the Philippines Los Baños
  • Jason R. Albia ⋅ PH Institute of Mathematical Sciences and Physics, University of the Philippines Los Baños
  • Ranzivelle Marianne L. Roxas-Villanueva ⋅ PH Institute of Mathematical Sciences and Physics, University of the Philippines Los Baños

Abstract

Crop disease identification through image analysis is proven to be more cost effective and less time consuming than traditional approaches. In this study, deep learning is employed to predict the occurrence of Cassava Phytoplasma Disease (CPD). A pre-trained Convolutional Neural Network (CNN) architecture, Inception-ResNet-V2, was utilized on a cassava image dataset through transfer learning. The model has 83.78% accuracy, 80.95% precision, 89.47% recall, and 0.836 area under the curve (AUC). This rapid disease identification would be beneficial to farmers as they can isolate CPD-infected crops at a much earlier time.

Downloads

Issue

Article ID

SPP-2021-PB-17

Section

Poster Session B (Complex Systems, Photonics, and Interdisciplinary Topics)

Published

2021-10-05

How to Cite

[1]
AJO Caringal, MP Martinez, SJB Manaday, MRJB Martirez, GAS Burgonio, MD Ebuenga, JR Albia, and RML Roxas-Villanueva, Deep learning approach for Cassava Phytoplasma Disease image classification, Proceedings of the Samahang Pisika ng Pilipinas 39, SPP-2021-PB-17 (2021). URL: https://proceedings.spp-online.org/article/view/SPP-2021-PB-17.