Easy Agriculture: Crops’ disease detection, pesticide, fertilizer and crop recommendations

Abstract

Around the world due to pests and pathogens almost 50% of the agricultural produce is lost which is so alarming given the fact that many people die everyday due to starvation in poor nations. Crop diseases disturb the normal growth and physiological processes. It is estimated that every year 20-40% of crop loss is reported and, in some cases, whole production gets destroyed. So, to produce higher yield and for sustainable agriculture it is important to identify any diseases from the early stage itself. Technology can do a great help in this cause to detect plant disease by using various AI techniques. It is also important to recommend proper pesticides for the persisting disease. The model proposed is based upon a 9 layer resnet deep learning algorithm that takes in present time images of various crops and detects the disease & also recommends the suitable pesticide. Plant Village Dataset taken from Kaggle comprising 87000 images (38 Classes,13 Crops) is used. A custom dataset is also built consisting of disease-description-measures to be taken-pesticide or fertilizer to be used. The end system developed also has two other models integrated that are used for crop and fertilizer recommendations. They are built using the Random Forest Classifier algorithm and a parameter conditional statements function.