Detection of Tomato Leaf Disease Using Deep Learning and Computer Vision

Abstract

Tomato is a common food product consumed across the world. It has been used in preparing several meals around the world. There have been issues around its price changes due to the impact of diseases in its production. Farmers have had to spend so much to address the issue, hence, posing a significant threat to agricultural productivity. This has also led to losses for farmers. An efficient and more accurate detection system is hereby essential. In this research, a hybrid approach for detecting tomato leaf disease using a deep learning model and computer vision is proposed. This leverages the power of a convolutional neural network model, Efficientnet_b0, a subset of the EfficientNet model, for learning and extracting features from tomato plant images. This deep learning model was trained on a diverse dataset of annotated images representing various tomato diseases and healthy plants. The trained model demonstrates its ability to accurately classify and identify different disease manifestations in real-time using the plant village dataset from Kaggle. Computer Vision techniques are incorporated to enhance the system's overall performance. The approach returned an accuracy of 0.97 with a real-time deployment potential, enabling farmers to diagnose diseases in their fields early.