Deep Learning-Based Identification of Arecanut Plant Disease in the Tumkur Agricultural Zone

Abstract

Arecanut cultivation in the Tumkur agricultural belt faces notable yield losses due to the prevalence of leaf diseases, underscoring the need for a reliable and automated detection system. This study proposes a deep learning driven solution utilizing transfer learning to classify arecanut leaf diseases from a carefully curated image dataset. Five convolutional neural network (CNN) architectures DenseNet-201 (88.69%), ResNet-50 (95.80%), EfficientNet-B0 (87.36%), Xception (88.41%), and ResNet-20 (94.92%) were trained and evaluated using perfor­mance metrics including accuracy, precision, recall, F1-score, and confusion matrix analysis. The ResNet-50 achieved the top accuracy of 95.80%, demonstrating superior feature extraction capabilities and consistent results across multiple disease cate­gories. These outcomes highlight the promise of ResNet-50–based deep learning frameworks in enhancing precision agriculture and strengthening disease management in arecanut farming.