Diagnosing Diseases in the Leaves of Food Crop Plants Using Residual Network – Powered by Microsoft Azure Custom Vision Service
Millicent Ama Agyir
Department of Computer Science and Informatics, University of Energy & Natural Resources, Sunyani, Ghana.
Eric Ayintareba Akolgo
Department of Computer Science, Regentropfen College of Applied Sciences, Bolgatanga, Ghana.
Dennis Redeemer Korda *
Department of Computing and IT, Bolgatanga Technical University, Ghana.
Adebayo Felix Adekoya
Department of Computer Science and Informatics, University of Energy & Natural Resources, Sunyani, Ghana.
Obeng Owusu-Boateng
Department of Mathematics and ICT, Bimbilla College of Education, Bimbilla, Ghana.
*Author to whom correspondence should be addressed.
Abstract
Background: The purpose for this research lies in the growing necessity for efficient and accurate plant disease identification methods to safeguard food crop production. Leveraging Azure Custom Vision Service and Residual Networks (ResNet) addresses the challenges of traditional methods by providing a cost-effective, user-friendly, and high-performing solution for classifying plant leaf diseases.
Methods: Azure Custom Vision Service was selected for its user-friendliness, cost-effectiveness, and ability to rapidly train and test image classification models. According to the literature, Azure Custom Vision has been effective in various tasks such as object detection, image segmentation, and classification, making it suitable for this study's goal of classifying plant leaf diseases. This research employed Residual Networks (ResNet) for their superior performance in image classification and ability to handle vanishing gradient problems, a common issue in deep neural networks, thus enhancing the accuracy of disease classification in food crop plants.
Findings: In this research, a customized model for classifying common diseases in the leaves of food crop plants has been developed using the Azure Custom Vision AI platform, based on ResNet algorithm. Based on this, it can be concluded that the first specific objective, which was to apply Residual Network (ResNet) algorithm to analyze the pattern of food crop plant diseases and predict the diseases by using Microsoft Azure Custom Service as a toolkit, has been achieved. The model was inference in a plant disease diagnosis software which allows users to upload an image from their computer, and within a maximum of twenty (20 seconds), the software can identify the disease and recommend some treatments. This also indicates that the second specific objective of the study; to develop a user-friendly and efficient interface that improves the ease of use of the software system for farmers and agronomists has been achieved.
Novelty and Applications: This proposed system is novel in its approach to using Azure Custom Vision Service based on ResNet for classifying diseases in food crop plants. Additionally, the system's performance surpasses that of existing state-of-the-art methods, further highlighting the novelty and effectiveness of the proposed model.
Keywords: Residual network, Microsoft azure, food crop plants, pathological diseases, image classification, machine learning