A transfer learning-based approach for the classification of tomato leaf diseases using modified classification base

Diseases affecting tomato leaves represent a significant risk to agricultural yield and quality, making swift and accurate identification essential for sustainable farming and reducing reliance on herbicides.Traditional manual evaluation methods are labor-intensive, subject to bias, and more likely to be erroneous.Deep learning, particularly via transfer learning (TL), has revolutionized plant disease detection by providing automated and highly precise classification.This study introduces a TL based customized classification model that classifies tomato leaf diseases into four unique categories namely Early Blight, Late Blight, Yellow Curl Leaf Disease, and Healthy leaves.The model is developed utilizing a diverse collection of accurately labelled images of both healthy and diseased tomato leaves sourced from the Plant Village dataset, a renowned and high-caliber dataset available on Kaggle.To improve performance, data augmentation techniques (such as rotation, flipping, brightness, and contrast modifications) are utilized, enhancing robustness and reducing overfitting.The effectiveness of the model is evaluated using metrics like accuracy, precision, recall, F1-score, and confusion matrix analysis, illustrating its superior classification performance compared to conventional machine learning methods.The results shows that the customization done in the classification part of the popular deep learning-based architectures namely VGG16, VGG19, ResNet50, InceptionV3, AlexNet and DenseNet, for the classification of diseases achieves comparable accuracy.