Disease Detection in Hibiscus Plant Leaves: A CNN-SVM Hybrid Approach

The presented article abstract summarizes the findings of a study that used CNN and SVM methods for identifying disease in Hibiscus plant leaves. The effectiveness of the categorization model throughout several illness categories was thoroughly analyzed, producing crucial Precision, Recall, and F1-Score parameters. The analysis showed remarkable Precision scores between 39.71% and 85.40%, confirming the model's precision in recognizing genuine positive cases throughout each illness category. Accordingly, Recall scores ranging from 52.35% to 86.00% show how well the model can identify real positive cases. F1-Scores, which range from 48.99% to 84.90%, also represent the model's proportion of sensitivity and precision. Macro, Weighted, and Micro Averages are included in the evaluation to account for distributional differences and to capture the general outcome across every class. This study promotes the creation of reliable disease forecasting algorithms by combining the strengths of CNN and SVM techniques. It also provides insightful information about how to use these models to detect leaf illnesses in Hibiscus plants.

I. Introduction

As the backbone of human society, farming is essential to providing sufficient nourishment and supporting economies all over the world. However, a variety of biotic and abiotic factors, such as diseases of plants, continually place food production in danger. Because of its aesthetic significance and usage for ancient medical care, hibiscus (Hibiscus rosa-sinensis), widely known as Hibiscus in various countries, possesses both aesthetic and agricultural significance. The state of the foliage on a plant is one of the key markers of that plant's health. Photosynthesis and transpiration are made possible by branches, which act as crucial contact between the plant and its surroundings. But they are also prone to a wide range of illnesses brought on by bacteria, fungi, infections, and other pathogens. Early disease detection is essential to stopping its propagation and reducing the harm it causes to agricultural quality and output. The subject of identifying plant diseases has seen a radical transformation recently because of developments in neural networks and algorithms for learning. Traditional detection techniques frequently rely on visual examination by agronomists and specialists, which can be subjective, will laborious, and scope-restrictive. The production of precise and effective machinery for disease diagnosis and categorization has been made possible by the incorporation of computational methodologies. Convolutional Neural Networks (CNNs) have become the industry standard for image analysis, achieving outstanding results in a wide range of tasks like recognizing things, segmentation of imagery, and medical evaluation. CNNs may capture subtle designs and materials that could be invisible to the viewers by utilizing their hierarchical structure. Support-vector machines (SVMs), on the contrary hand, are a traditional machine learning method renowned for their sturdiness when dealing with high-dimensional data and their efficacy in binary grouping tasks. SVMs are able to distinguish between various plant-associated disease states by locating a hyperplane that takes advantage of the margin within classes. In order to predict various illnesses in the leaves of the Hibiscus plant, the present work gives a thorough investigation of the combination of CNN and SVM techniques. Utilizing the advantages of both strategies, these methodologies work in harmony to increase the accuracy and predictability of illnesses diagnosis. This investigation aims to support the growing body of research in the area of automatic plant illness diagnostics by combining the feature acquisition skills of CNNs alongside the selective capacity of SVMs.