Integration of CNN and Random Forest Classifier for Detecting Tomato Diseases
TL;DR
This work proposes Convolutional Neural Network with random forest (RF) (CNN-RF) trained on a single plant which is tomato, which aims to solve the problem by using CNN and Random forest classifier to leverage machine learning algorithms trained on extensive datasets of tomato leaves to identify symptoms and patterns indicative of diseases at early stage.
Abstract
Plant diseases, particularly in tomato crops pose a significant threat to agricultural productivity which results in yield losses, accounting for an estimated 10-30% of global tomato production annually. In this era characterized by technological advancement farmers continue to follow traditional practices regarding disease identification in crops. Rather than depending on the modern specialized tools, they heavily rely on personally and visually inspecting the crops to detect any sign of disease, this method is solely based on the farmers expertise and this presents several challenges and limitations in agricultural research and is prone to human error which in the worst scenario the undetected crop infection can cause the entire crop to decline. Many of the existing systems that are available usually tend to have been trained on multiple plants and as a result they are not very effective. Therefore, this work proposes Convolutional Neural Network (CNN) with random forest (RF) (CNN-RF) trained on a single plant which is tomato. The model will be trained and validated using a dataset of over 18,000 labeled images of tomato leaves having multiple disease classes. Using this CNN and Random Forest (RF), this system aims to solve the problem by using CNN and Random forest classifier to leverage machine learning algorithms trained on extensive datasets of tomato leaves to identify symptoms and patterns indicative of diseases at early stage.