Banana leaf diseases, including Sigatoka, Pestalotiopsis, and Cordana, pose a significant threat to banana crops, affecting both leaf and fruit quality. Early detection of these diseases is essential for implementing effective control measures and ensuring crop health. In this study, we developed a deep learning-enabled multimodal framework for the automated detection and classification of banana leaf diseases in precision agriculture. The research applied the CRISP-DM methodology, which consists of six phases: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. A dataset consisting of 937 RGB images, divided into 4 distinct classes, was utilized for training. Four deep learning algorithms—Convolutional Neural Networks (CNN), EfficientNet, ResNet, and VGG19—were trained on this dataset. Given the high computational demands of CNN, ResNet, and VGG19, EfficientNet was selected for model training due to its efficiency and performance. TensorFlow was employed for model training and performance optimization. The results demonstrate the potential of deep learning models for accurately detecting and classifying banana leaf diseases, offering a promising approach to enhance crop management in precision agriculture.

Keywords: Banana Leaf Diseases, Sigatoka, Pestalotiopsis, Cordana, Deep Learning, CNN, EfficientNet, ResNet, VGG19, Precision Agriculture, Automated Detection, CRISP-DM, TensorFlow, Crop Disease Detection, Disease Classification, Agricultural AI.