Classification of Costal Area Diseases of Arecanut using Dual Convolutional Neural Network

This study presents an automated approach for the detection and classification of coastal area arecanut diseases using Dual Convolutional Neural Networks (DCNNs). A custom dataset consisting of 1,000 images, captured under expert supervision from Navilgon village in Honnavara taluk, Karnataka. This was developed to represent four classes like Healthy, Rot, Split, and Rot+Split. All these images were preprocessed by resizing them to 128×128 pixels and converted into numerical arrays to facilitate model training. The proposed DCNN model incorporates double convolutional blocks, batch normalization, dropout, and max-pooling layers to efficiently extract and learn both lower-level and upper-level features. The architecture culminates in a softmax-activated dense layer to classify the input images into one of the four disease classes. The dataset was split in an 80:20 ratio for training and testing, and the model was trained using the ReLU activation function, Adam optimizer, and categorical cross-entropy as the loss function. Confusion matrix shows Healthy and Rot classes have the highest accuracy of 80%. Similarly Split and Rot + Split are slightly lower accuracy of 65%.