DGRLeafNet: multimodal fusion machine learning model for mango leaf disease classification

The diseases affecting mango leaves is the main issue in global mango cultivation, which results in reductions in both production and quality. Accurate identification is very important for diagnosing these diseases. The primary objective of the proposed model is to develop accurate and robust machine learning model to automate the classification of mango leaf diseases at a very early stage. This study presents a hybrid deep learning model that integrates ResNet152 and DenseNet121 for thorough classification of mango leaf illnesses. The deep residual connections in ResNet152 enable the effective training of a highly deep network, while the dense connectivity in DenseNet121 promotes better feature reuse and gradient flow. To find optimum values of hyper parameters of the framework, a Genetic Algorithm (GA) is employed. The proposed approach is evaluated using a carefully selected dataset of mango leaf images, which encompasses different disease types along with healthy samples. Experimental findings indicate that the hybrid ResNet-DenseNet model, optimized through GA, achieves improved accuracy and generalization when compared to separate models, positioning it as an effective tool for automated detection of mango leaf diseases in precision agriculture. The dataset contains images mango leaf diseases, including Anthracnose, Powdery Mildew, and Bacterial Canker with respective biochemical data The results proves that the hybrid model enhances metrics such as precision, recall, and overall accuracy when compared with the individual models.