The Banana Leaf Disease Classification System is a robust deep learning-based application designed to efficiently identify, classify, and predict various pathogens in banana crops using advanced computer vision and neural network principles. It follows a modular pipeline architecture, where input images interact through an intuitive preprocessing stage that facilitates noise reduction, normalization, and high-level feature extraction. On the backend, the system employs a Convolutional Neural Network (CNN) to analyse visual symptoms of critical diseases such as Black Sigatoka, Panama Wilt, and Banana Bunchy Top Virus. The use of the Adam optimizer and categorical cross-entropy ensures efficient training and accurate convergence, while data augmentation techniques protect against overfitting and improve model generalization. Furthermore, a comparative analysis against traditional machine learning algorithms like SVM and Random Forest is implemented to validate the system’s superior performance, ensuring that agricultural stakeholders receive high-accuracy diagnostic results. The integration of SoftMax activation and probability-based classification enables quick and reliable identification,
Ranjini et al. (Wed,) studied this question.