Deep Transfer Learning Approach to Detect Dragon Tree Disease

Dragon fruit cultivation in Bangladesh is steadily growing. Farmers are increasingly interested in planting dragon fruit in cropland at the marginal level. The cost is little and the yield is simply achieved with great efficiency. With the rising output of the fruit, its cultivation is likewise steadily increasing. Consequently, a multitude of ailments affecting the dragon tree are emerging. A significant number of trees are perishing prior to the growth of dragon fruit. Hence, we offer a sophisticated deep-learning model capable of accurately detecting and classifying different diseases affecting dragon fruit. For this particular instance, we utilized 6000 photographs that were captured manually using our mobile device. We have identified three prevalent diseases that affect dragon fruits: Anthracnose, Root rot, and Stem canker. I have also acquired a recently harvested dragon fruit. Four distinct models have been selected in this instance. These options are Restnet50, VGG16, CNN, and VGG19. All the models yielded satisfactory results, however, VGG19 exhibited exceptional performance. The VGG19 model achieved an accuracy of 92%. This is an excellent outcome. The device successfully identified three diseases that impact dragon fruit trees and was also capable of detecting freshly harvested dragon fruit. The practical application of this technology will result in a transformation within the community of dragon farmers. Implementing this measure will enhance farmers’ crop production while reducing mortality rates among dragon trees.