Blight Disease Detection System using Deep Learning

Abstract

Plant diseases can have a significant negative impact on the yield of both subsistence and commercial farmers, potentially leading to crop failure and financial loss. Traditional methods for detecting plant diseases are often time-consuming, costly, and require specialized skills. However, recent advancements in computer vision and deep learning technologies have enabled the creation of automated systems for detecting plant diseases that are both speedy and accurate. This study utilizes an open-source dataset from Kaggle, comprising 4082 images of plant leaves, along with advanced deep learning algorithms, specifically Convolutional Neural Network (CNN) and Graph Neural Network (GNN) to detect diseases without requiring professional expertise. The dataset was split into training and testing sets (ratio 80:20). CNN was employed to automatically extract features from the image data, effectively identifying patterns and details that are imperceptible to the human eye. GNN was used to model the relationships and dependencies between different plant conditions, enhancing the system's capability to understand the spread and severity of diseases. The plant disease detection system demonstrated exceptional performance, boasting an accuracy rate of 97%. Additionally, the system achieved a remarkably low loss value of 0.03, indicating the effectiveness of its predictive capabilities. These results underscore the reliability and precision of the system in diagnosing plant diseases, offering farmers valuable insights to safeguard their crops and optimize agricultural productivity.