Plant Disease Detection using Deep Learning and Convolutionary Neural Network

Abstract

Modern agriculture has evolved into much more than a means of feeding ever-increasing populations. Agriculture affects a country's economy in some way. With the population growing by the day, the primary sector must be given more attention. According to a World Bank report, three out of every four people in developing countries live in rural regions and earn less than Rs.200 per day. Agricultural progress is required to improve the quality of agro-based industry goods, particularly in emerging countries. As a result, early detection of plant diseases could be crucial in preventing agricultural losses. Plant disease detection is based on the principle that all information that aids people in growing food should be freely available to everyone on the earth. The next important disease diagnostics tool that aids in making a vision of productive agriculture into a reality is developing algorithms that can reliably detect a disease based on an image. The goal of this project is to develop an artificial intelligence software that can identify and classify plant illnesses. We'll be using PlantVillage, a public dataset of 54,444 photos, and PyTorch as our deep learning platform. We have worked for Potato and Pepper plant leaves, but the approach can be used for any plant and dataset can be extended accordingly anytime. After Data collection, model have been built using tensorflow, CNN, data augmentation is done by using tf dataset. Tf server will built backend server using FastAPI and then we can also deploy our server Model to Google cloud (GCP). Also the Google cloud function that is running in front end and these functions will be invoked by mobile application return in react native so it is an end to end project. For optimization we have used quantization process Tensorflow lite. Frontend is by using ReactJs and react native deployment is by using GCP (Google Cloud Platform) and GCF (Google Cloud Functions). Plant diseases will be detected using photos of plant leaves. As a result, we feel that early diagnosis of plant diseases will undoubtedly aid in the maintenance of agricultural stability and the advancement of a country's growth.