A tomato leaf fungal disease image dataset and a metaheuristic-based framework for optimizing machine learning classification

This study introduces a new RGB dataset specifically developed to detect tomato fungal disease and presents a metaheuristic-based machine learning framework for disease classification. The dataset comprises 3,000 labeled images evenly distributed across five categories: four major tomato fungal diseases with common symptoms ( Alternaria alternata, Alternaria solani, Botrytis cinerea, Fusarium oxysporum ) and healthy leaves. After extracting color, texture, and shape features from the images, a genetic algorithm (GA) was employed for two key tasks: feature selection and hyperparameter optimization of two machine learning models (support vector machine-SVM and random forest-RF). GA-driven optimization led to substantial improvements in classification performance, with overall accuracies increasing from 68.17% to 87.17% for SVM and from 80.83% to 98% for RF. The RF model achieved high precision across all categories, exceeding 95% in most cases. Additionally, RF demonstrated consistent improvements in precision, sensitivity, and specificity for all disease categories. These results highlight the effectiveness of GA in enhancing classification performance, particularly for complex datasets with overlapping visual symptoms. Feature selection analysis revealed that texture features were the most influential for classification, followed by color and shape features. The proposed framework demonstrates the potential of combining structured datasets with intelligent optimization techniques for real-time plant disease diagnosis, supporting the development of smart agriculture technologies and sustainable crop management practices.