ACCURATE AND SCALABLE TOMATO-LEAF DISEASE DETECTION USING HYBRID RESNET50–VGG16 MODEL FOR AGRICULTURE

Abstract

Tomato leaf diseases pose a severe danger to agricultural productivity and global food security. The automated tomato leaf disease detection system presented in this study is based on a hybrid deep learning model that combines the feature extraction capabilities of the VGG16 and ResNet50 architectures. The system was tested on a publicly available dataset of tomato leaf images against 4 baseline models: CNN, ResNet50, VGG16, and MobileNetV2. With a 97.99% accuracy, a 98.04% precision score, 97.99% recall score, and a 97.98% F1-Score, the suggested hybrid model performs better than individual models. This was verified after extensive trials with the datasets. The model's resilience and efficacy in reliably diagnosing a variety of tomato leaf diseases were further validated by visual studies such as confusion matrices, bar graphs, radar charts, and heatmaps. Since the model can capture both fine-grained textures and deep hierarchical characteristics, the hybrid model performs better, providing more thorough pattern identification. Future research should concentrate on improving the system for edge devices, extending to multi-crop disease detection, and integrating explainable AI approaches. The suggested approach presents a viable first step toward automated, scalable, and accurate plant disease monitoring, which will enhance crop management, lower the need for pesticides, and promote sustainable farming methods.