Tomato Leaf Disease Detection Using Transfer Learning with EfficientNetB3: A CNN-Based Approach
DOI:
https://doi.org/10.4314/cajost.v7i3.18Keywords:
Machine Learning, , Tomato leaf images, , Convolutional Neural Networks, , Plant diseases , and Disease detection System.Abstract
Crop diseases are a real threat to our food supply, and catching them early is
crucial to protect plants and boost harvests. Many farmers struggle to spot and
manage these diseases in time, which can lead to significant losses. Our study
tackles this challenge by developing a CNN-based Image Tomato Disease
Detection System to help farmers quickly and accurately identify tomato leaf
diseases. Using a powerful convolutional neural network called EfficientNetB3,
our system analyzes tomato leaf images to sort them into three categories:
healthy, yellow leaf curl virus (YLCV), or bacterial spot (BS). We trained the
model on 1,273 tomato leaf images, running tests for 10 and 20 rounds (or
epochs), achieving accuracies of 33% and 58%, respectively. After 20 rounds,
the system also showed a precision of 52%, recall of 54%, and F1-score of
53%, proving it gets better with more training. This automated tool offers a
practical way to spot tomato diseases early, helping farmers protect their crops
and improve both yield and quality.
Downloads
