Plant Doctor

Abstract

Agriculture is the foundation of all civilizations and cultures. According to the Food and Agricultural Organization (FAO), agriculture is essential to the global economy because it provides food for more than half of the world's population (62 percent). But because of crop losses, more than 40% of cultivated plants become unusable annually. In Pakistan, agriculture is the primary source of income for most of the population. Plant diseases are difficult to detect with the naked eye most of the time. PlantDoctor is a cross-platform mobile application that uses deep learning to identify plant diseases by analysing plant leaves. PlantDoctor is a cross-platform mobile application which consist of 3 main components that a cross-platform application, a model which is trained by deep learning for detecting the disease, and a server-side application that works as the API gateway for the whole application. The report consists of the research part by observing and distinguishing existing similar systems and case studies. Deep learning has been used as an automatic crop disease detection. It was proposed to use Convolutional Neural Network as the deep learning algorithm but the results we got were not good enough. So, we had to move on to Transformers for tetter results. By using Transformers, the deep learning model has achieved 98.75% accuracy and it was performing well on real-time data as well as compared to CNN other algorithms like using Visual Geometry Group (VGG) and Residual Neural Network (ResNet). It has completed the perspectives used to construct this proposed system, design stage, usage and capacities, and the most critical basic assessment within this application.