Deep Learning Precision Farming: Tomato Leaf Disease Detection by Transfer Learning

Abstract

Agriculture sector in India is facing few issues like low yield per hectare, traditional techniques of production, fertilizers and pesticides used at great scale. In this research, we implemented drones based precision farming system to effectively identify high disease area in farm using Convolution Neural Network. At later stage, we will spray precise pesticides on affected area based on infection level with the help of drones. The dataset contains 2100 images of tomato leaves from internet and 500 images from local farms. Transfer learning is used to retrain the inception model of Google to train CNN for tomato leaves. Leaves are classified in 3 broad category as good, average and bad for pesticide intensity. 99% accuracy is achieved when training percentage is increased to 85%. Execution speed of system is also fast as we are using transfer learning on inception model.

Keywords: Precision Farming, Tomato Leaf Disease, Deep Learning, Tensor Flow, CNN, Transfer Learning, Drone