A Comprehensive Study of Application Domains of IoT and AI in Horticulture: A Case Study of Plant Leaf Disease Detection and Classification using CNN

Abstract

Horticulture is a branch of agriculture that deals with growing fruits, nuts, vegetables, flowers, and ornamental plants. It is one of the fastest-growing economic sectors contributing 30% to India's Gross Domestic Product (GDP), upholding India as the second largest horticulture producer in the world. However, Horticulture crops face many challenges including effective utilization of land, shortage of labor, shortage of water, low soil fertility, early detection and treatment of plant disease, pest control, crop monitoring, timely harvesting, and yield prediction. In this paper, we presented a qualitative survey addressing the mentioned challenges in agriculture and horticulture using technologies like IoT and AI. We discussed the application of IoT at various stages of farming, from irrigation to crop harvesting, and explored the application of AI techniques for disease detection, yield prediction, etc. We also presented a case study of plant leaf disease detection and classification using Convolutional Neural Network (CNN), an application of AI in agriculture. We used transfer learning (TL) based CNN model to train and validate using an enormous dataset of 87,867 labelled images. We were able to achieve training and validation accuracies of 0.69(69%), and 0.86 (86%) respectively, for 10 epochs, 0.0001 learning rate, and 50 % dropout rate. The model demonstrated training and validation losses of 1.05 and 0.53, respectively.