Early Detection of Plant Leaf Disease Using Convolutional Neural Networks

摘要

Plant disease is a continuing challenge for smallholder farmers, which has an impact on income and food production. Identifying the disease at starting stage and preventing it from spreading to other parts of the plant is a challenge even for the experts in the field. Experts can identify and diagnose the disease, but the process is time-consuming and a good observation of the infected part is needed. However, these experts are not available readily to smallholder farmers, who are a major part of our country. An adequate method is therefore required to detect plant leaf diseases at starting stage. The recent revolution in smartphone perforation and advancement in computer vision models has provided a way for computer vision applications in the agriculture. Convolutional Neural Networks (CNN) considered as state of the art in classification of images and have the ability to produce a conclusive diagnosis.In this article, a Transfer Learning approach is used, in which a pre trained model is used to train on pictures of different fruit plant leaves from the Plant Village dataset, covering various diseases as well as safe samples.