Title: Detecting diseases in plant leaves: an optimised deep-learning convolutional neural network approach

Authors: Saraansh Baranwal; Anuja Arora; Siddhant Khandelwal

Addresses: Jaypee Institute of Information Technology, A-10 Sector 62, Noida – 201310, India ' Jaypee Institute of Information Technology, A-10 Sector 62, Noida – 201310, India ' Jaypee Institute of Information Technology, A-10 Sector 62, Noida – 201310, India

Abstract: A country's economy relies heavily on its agricultural productivity. Most of the crops consumed daily by the population are prone to diseases. Identifying and preventing the disease at an early stage is a challenge even for the expert's eye. Therefore, an appropriate system is needed to detect plant disease in its initial stages. This paper employs the approach of convolutional neural networks to automatically detect and address the issue. Images for various plants covering 35 different classes of plant diseases and a total of 29,180 RGB images of diseased and healthy plant leaves has been used for the purpose. Image filtering, compression, and data generation techniques have been used to further increase the training dataset and achieve high accuracy across all classes. Net accuracy of the model reaches a 98.34% success rate on the entire dataset. The data has been sampled and generated from severely downsized images, with a significant focus on speed as well as accuracy.

Keywords: plant leaf; disease detection; leaf images; machine learning; convolutional neural network; CNN; LeNet; Keras.

DOI: 10.1504/IJESD.2021.114562

International Journal of Environment and Sustainable Development, 2021 Vol.20 No.2, pp.166 - 188

Accepted: 30 May 2020
Published online: 27 Apr 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article