Investigation on paddy crop disease detection and classification using SVM and CNN Algorithms

Abstract

Rice cultivation plays a fundamental role in ensuring food security and sustaining agricultural economies, particularly 
in developing countries. One of the major factors affecting rice productivity is the occurrence of leaf diseases, which 
significantly reduce crop yield when not identified at an early stage. Conventional disease diagnosis relies on visual 
inspection by farmers or agricultural experts, making the process time-consuming, subjective, and often inaccurate. 
Although deep learning techniques have shown promising results in automated disease recognition, many standalone 
convolutional neural network (CNN) models experience performance degradation when dealing with visually similar 
disease categories and complex field conditions. 
To address these challenges, this paper presents PaddyCare AI, an intelligent paddy leaf disease diagnosis system 
based on a hybrid deep learning architecture. The proposed framework employs the pre-trained MobileNetV2 network 
to perform efficient feature extraction while utilizing a Support Vector Machine (SVM) classifier to improve decision 
boundaries and classification performance. The model was trained and evaluated using a balanced dataset 
comprising five classes: Bacterial Blight, Blast, Brown Spot, Tungro, and Healthy leaves. 
Experimental results demonstrate that the proposed MobileNetV2–SVM hybrid framework achieves superior 
classification accuracy compared with conventional CNN-based classifiers while maintaining computational 
efficiency suitable for practical deployment. To enhance accessibility, the trained model is integrated into a web-based 
application that enables farmers to upload leaf images and receive immediate disease diagnosis along with preventive 
recommendations. The proposed system facilitates timely intervention, minimizes crop losses, improves agricultural 
productivity, and promotes precision farming through artificial intelligence.