Crop Disease Detection Using Computer Vision

Abstract

Historically, Bangladesh is an agriculture-based country. Approximately 40% of its population is involved in the agriculture sector. Although our country has achieved food sufficiency in recent years and many of the obstacles in farming has been removed due to the emerging of modern technology, the farmers in the rural area are still suffering from multiple issues that are harming the harvesting of the crops. The various diseases of the plants and crops are among them. A tremendous amount of product damage due to diseases. It has a bad impact on the farmer as well as a bad effect on our economy. This occurs due to not to able to detect disease in time. If the disease can detect at the earliest time, it is possible to decrease the damage rate. It is tough and arduous to detect these diseases manually. In this project, we have taken various methodologies to identify disease. We have gathered a database based on different images of disease of potato and tomato to train and test. We have used PyTorch library which has developed by Facebook deep learning team. We also have used Torchvision for image segmentation and ResNet 152 for training our dataset. It helps to identify disease type. This model has a precise excessive perspective to be further enhanced in the future.