Grape Disease Detection using Drone Imaging and Deep Learning – A Pilot Study

Early-stage detection of the diseases is necessary for maintaining crop health and maximizing agriculture productivity. Traditionally, monitoring of outbreaks is often through intensive labor in the field and subjective human judgment. This study demonstrates drone technology as a means of improving disease diagnosis by the remote sensing technique embedded with machine learning and image processing techniques. Cameras mounted on the drones capture images of the vineyard from above to identify disease symptoms such as color variation, textural variation and abnormal growth forms. Powdery mildew is one of the grapes diseases which reduce the yield of farmers. The highest incidence of 72.55 percent at 35 to 45 days after pruning has been reported at Madhampatty village in Coimbatore district. For identifying the incidence of grape diseases, deep learning techniques are used, and it helps us for early detection. A comparative study of different deep learning models for grape health classification is experimented. DenseNet and AlexNet achieves the highest accuracy of 99.88% and 100% in two different datasets and strong performance metrics. ResNet 50 and InceptionV3 also showed good results. The findings highlight the need for selecting the best model based on accuracy, precision, recall, and generalization for agricultural monitoring.