An Advanced Cloud-Based Plant Health Detection System Based on Deep Learning
Syed Hauider Abbas, Swati Vashisht, Garima Bhardwaj, Ruchira Rawat, Anurag Shrivastava, Kavita Rani
Abstract
Agricultural diseases can hinder plant growth, which ultimately causes crop death. Farmers traditionally identify diseases by speaking with regional experts and conducting additional experiments. Farmers may find it challenging to comprehend the current situation given the prevalence of new diseases. India suffers 15% of the world’s annual agricultural losses, which is significant. To stop further crop damage, automated disease detection using machine learning and the cloud is crucial. The existing machine learning models in this proposal lacked sophistication, reliability, portability, and ease of use in some respects, making it cumbersome for others to actually examine the results in real time. In this proposal, by using Amazon’s cloud service AWS (Amazon web Services) to implement the proposal and host the website via the cloud, a farmer can take a picture of a leaf and send it to the users of the website. It can apply it to interface uploads to learn how it works. Using MobileNet, a convolutional neural network (CNN), and machine learning libraries such as TensorFlow and OpenCV to detect plant diseases with an average accuracy of 98.78% using a data set of approximately 2000 plant leaf images identified in real time.