Crop Disease Classification Model using Convolution Neural Network

Abstract

Agriculture is the backbone of food security and a key to newlineeconomic stability around the globe. Soybean is a major crop newlinewith high nutritive value, which is widely used in food, feed, and newlineindustrial applications. Soybean plants are also highly vulnerable newlineto diseases, which can result in significant yield losses. Timely newlineand precise disease detection is of utmost importance to newlineminimize these losses and to supplement crop management newlinepractices. newlineThe study demonstrates a deep learning-based AI solution for newlineclassification of soybean diseases. As CNN can extract newlinehierarchical features which are clear more powerful in the image newlinebased classification tasks. newlineProposed research deployed several deep learning algorithms, newlinesuch as EfficientNet-B0, MobileNet V2, RestNet-50, and custom newlineCNN architectures for soybean crop disease classification. We newlineExperiments were performed on three color spaces of images: newlineGray, HSV, and RGB. Results show that our RGB images based newlineon CNN model has the highest classification accuracy (98.60%), newlineindicating the important disease features captured by our data. newlineThe custom CNN model trained with HSV images showed great newlineperformance as a versatile model for complex disease symptoms newlinelike Vein Necrosis, Dry image. CNN with Gray image required low newlineresource utilization like CPU, and RAM Utilization. newlineAmong all proposed models based on transfer learning, newlineEfficientNet-B0 achieved an accuracy of 96.44% which validated newlineits efficacy in classifying soybean diseases. For mobile and edge newlineapplications, MobileNet V2 was the best option due to being the newlineleast resource intensive, making it available to low-space newlinedevices.This research demonstrates the necessity for AI-assisted newlinetechnologies aimed at precision agriculture and disease control. newlineThis research will help the plant production industry adopt newlineeffective measures for efficient and scalable plant disease newlineclassification systems.