Journal of Experimental Agriculture International

A disease and pest advisory system provides farmers with timely, scientific and location-specific support for identifying crop health problems and selecting suitable management actions. This study developed an artificial intelligence-based advisory system for okra diseases and pests using field-acquired images and farmer-oriented recommendation delivery. An image dataset representing three major okra disease categories, namely Yellow Vein Mosaic Virus, Enation Leaf Curl Virus and Powdery Mildew, was prepared from images collected at the MVRS field of Anand Agricultural University. A vernacular-language recommendation database was created using AGRESCO recommendations of Anand Agricultural University to support farmer-friendly advisory output. A graphical user interface was designed to allow farmers to upload disease images, receive automated disease classification and access corresponding management recommendations. Six convolutional neural network architectures, namely ResNet101V2, InceptionV3, VGG19, Xception, DenseNet201 and MobileNetV3, were evaluated using transfer learning. Data augmentation techniques, including reflection, scaling, rotation and translation, were applied to reduce overfitting and improve model generalisation. Among the tested models, ResNet101V2 achieved the highest classification accuracy of 99.45%, followed by DenseNet201 and Xception, each with 98.63% accuracy. The system also incorporated web-based advisory output in a vernacular language and audio support. The findings indicate that CNN-based classification integrated with advisory delivery can support timely decision-making for okra disease management, although broader field validation is required before operational deployment.