Fusion of Color, Shape and Optimization Techniques for Tomato Plant Leaf Disease Detection Using Machine Learning

Abstract

Tomato leaf diseases stance a major threat to tomato cultivation, causing substantial yield losses and economic damage. Early and precise identification of leaf diseases is vital for appropriate intrusion and disease control. This study investigates the effectiveness of ML techniques MLP Neural Network in classifying tomato leaf diseases using shape and color features extracted from leaf images. Four different optimization algorithms (SGD, RMSProp, AMSGrad and Momentum SGD) were evaluated across three learning rates (0.5, 0.01 and 0.00001) for training a neural network model. The results demonstrated that RMSProp and AMSGrad outperformed SGD and Momentum SGD in terms of classification accuracy. RMSProp achieved the highest accuracy at a learning rate of 0.5, while AMSGrad excelled at both learning rates of 0.01 and 0.00001 with 95.00% accuracy. The proposed method using shape and color features and the AMSGrad optimizer can be an effective tool for early detection and control of tomato leaf diseases.