An Intelligent Computer Vision System for Multi-Crop Plant Disease Detection Using Convolutional Neural Networks

Abstract

Plant diseases are influenced by environmental and climatic changes and represent a major natural threat that disrupts plant growth and can lead to plant death from the early stages of seed development through maturity. Many plant disorders are difficult to detect without adequate expertise, and farmers and plantation growers often lack the professional knowledge and resources required for accurate disease identification and management. To address this challenge, this study developed a convolutional neural network (CNN) model for the detection and classification of 24 plant diseases across five major crop species: banana, corn, rice, potato, and tomato. The primary objective was to create an accurate and easy-to-use tool that enables farmers to rapidly identify plant diseases. The CNN model was comprehensively evaluated using multiple performance metrics and achieved an accuracy of 96%. To enhance accessibility and practical usability, the model was integrated into both web-based and mobile applications, allowing farmers to capture or upload images of diseased plants and receive instant disease predictions. This user-centred approach supports timely intervention and improved crop productivity. An image data generator was incorporated to further enhance model performance, increasing classification accuracy to 97% and improving disease recognition reliability. The system can detect plant diseases in less than two seconds, demonstrating strong potential for real-time field deployment. Overall, the proposed integrated solution provides an efficient, accurate, and accessible approach to plant disease detection, with the potential to advance precision agriculture, promote sustainable crop management, and strengthen food security through the application of intelligent technologies.