Detection of Crop Diseases using Deep Learning via Android Application

Abstract

Crops have ever been a primary source of food for humans as it provides us energy to carry out our everyday tasks. Every person requires food for his survival. During the early ages, the food requirement was far less as the human population was very less and sparse but after the fourth industrial revolution population explosion occurred due to which there was a sharp increase in population, and as a result, the food demand also spiked due to which shortage of food had occurred which still exists. This shortage was primarily caused due to two reasons-Increase in Agriculture destruction and a sharp increase in population. Deep learning has brought a new era of introducing intelligence to our artificial devices to imitate a task like humans without being programmed with pre-defined rules to do so. In this paper, we propose to integrate Deep learning to reduce the loss of crops due to crop infections caused by various microbes. We implement an Android solution operating in a mobile environment that integrates the Deep Learning Neural Network and provides an on-device image recognition of crop diseases. The deep learning model acquires an accuracy of 95% and is a modified MobileNetV2 model which is converted to a Siamese Network. This model is deployed as an Android Application with high performance and a higher accuracy while only consuming the resource of that device. Due to all the factors, this solution can be widely implemented due to its higher accuracy as well as it is cost-friendly.