Real-Time Plant Leaf Disease Detection Using Machine Learning – Open CV

Abstract

There is a major danger to agricultural productivity posed by tomato leaf diseases, which lead to
significant drops in production. Accurate and prompt diagnosis of disease is crucial for prompt intervention and
effective therapy. This research strengthens a Counting Sort-based feature engineering technique to provide a
distinct deep learning method for disease identification in tomato leaves. Redundancy in features and computing
complexity are common problems with traditional image-based methods for disease detection. We circumvent these
problems by taking use of Counting Sort's efficiency to preprocess tomato leaf pictures and extract relevant features.
Specifically, Counting Sort is used to examine the distribution of pixel intensities within certain color channels; this
allows for the identification of significant textural and color-based traits associated with different types of sickness.
To classify tomato leaves into early blight, late blight, and healthy disease groups, a Convolutional Neural Network
(CNN) architecture is trained using these traits. To evaluate the efficacy of the proposed approach, a publicly
available collection of tomato leaf photographs is used. Experimental findings show that traditional deep learning
methods relying on raw pixel data or general feature extraction techniques are slower and less accurate than the
Counting Sort augmented deep learning model. To improve agricultural practices and boost food security, this
technology offers a practical means of automatically and efficiently identifying tomato leaf diseases. In this context,