Comparative Analysis of VGG-16, ResNet50, and EfficientNet-B1 with Optimization Techniques for Crop Disease Detection

Abstract

This paper investigates the application of convolutional neural networks (CNNs) to enhance the identification of leaf diseases in Malaysian agriculture, specifically focusing on tomato and potato crops. The study utilizes pre-trained models VGG16, ResNet50, and EfficientNet-B1, employing transfer learning methods. The PlantVillage dataset, known for its comprehensive collection of annotated images of healthy and diseased plants, forms the basis for training and assessing these models. The research aims to develop a reliable AI system for accurate and efficient disease identification, contributing to sustainable agricultural practices and food security in Malaysia.