Title: Detection of crop disorder using deep learning

Authors: Vinita; Suma Dawn

Addresses: Jaypee Institute of Information Technology, Noida, Uttar Pradesh, India ' Jaypee Institute of Information Technology, Noida, Uttar Pradesh, India

Abstract: An estimated 14% of global yield is lost to plant diseases each year, causing suffering to billions of people. Plant pathology studies diseases, microbes and climatic conditions that lead to plant death. Temperature, pH, humidity and moisture can cause plant diseases. Chemical misuse, environmental imbalance and drug resistance can result from misdiagnosis. Diseases can be diagnosed by human scouting. Image analysis of plant leaves can help diagnose diseases automatically. Automated disease detection involves image selection, pre-processing, segmentation, augmented features and model prediction. Crop diseases can be detected and classified accurately by Deep Convolutional-Networks since a few years ago. This paper compares deep learning approaches for predicting healthy and diseased leaves from Mendley database. We suggest variations that improve classification accuracy. In this work for disease, Deep CNNs are implemented including ResNet-50, Mobilenet, Densenet121, EfficientnetB0 and the proposed approach. Over 99% accuracy was achieved in detecting various crop diseases.

Keywords: deep learning; crop disease detection; ResNet-50; Mobilenet; Densenet121; EfficientnetB0; image processing.

DOI: 10.1504/IJGUC.2024.136725

International Journal of Grid and Utility Computing, 2024 Vol.15 No.1, pp.65 - 74

Received: 03 Jan 2023
Accepted: 09 Jul 2023
Published online: 19 Feb 2024 *

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