Revolutionizing Tomato Agriculture: Leaf Disease Detection Using CNN and Its Variants

India, being a large agricultural market is considered to be one of the major producers of tomatoes in the world having high economic value. However, the volume and quality of tomato crop output are diminishing day by day owing to several factors that impact the crop productivity resulting in significant losses for the farmers. The escalating challenges in tomato agriculture, therefore, demand innovative solutions for the timely and accurate identification of plant diseases. Numerous works have been provided in the literature to address these problems, however attaining high accuracy is still a major challenge. To tackle these inadequacies, we proposed an analytical approach based on Convolutional Neural Networks with a major emphasis on ResNet50 and VGG16 architectures to enable better complex plant disease pattern detection. For our experiment analysis and evaluation, we exploited a labeled dataset obtained from Kaggle comprising 10,388 images with 10 different tomato leaf disease classes. Our experiment illustrated satisfactory plant detection accuracy. It also outperforms some of the methods in the literature and attains 99.63% and 94.48% training accuracy at 20 epochs for ResNet50 and VGG16 models respectively. The enhanced result of our approach is evident from the knowledge of transfer learning that imports pre-trained Resnet50 and VGG16 models with several data augmentation techniques.