Abstract
India produces 20 million metric tons of tomatoes annually, with 150,000 metric tons being exported to international markets. India ranks as the leading producer and exporter of tomatoes globally, and tomato farming has a significant contribution to India’s agricultural economy, with millions of farmers relying on tomato farming for their livelihood. Tomatoes are in high demand during the summer, but cultivating them at this time of year is challenging because the hot climate increases the susceptibility to numerous diseases. In this study, we collected 5250 images of tomato leaves suffering from seven distinct diseases. We identified regions of interest, such as leaves, using the YOLOv8 framework for object detection and passed these to the ResNet-50 model for classification. Using classification assessment metrics, the effectiveness of this framework was evaluated and compared with other deep learning architectures, such as CNN, AlexNet, VGG-19, and EfficientNetV2B7. We provide optimal treatment recommendations based on disease identification using a comprehensive disease treatment dataset, along with detailed explanations through GPT-3.5. The results demonstrate that YOLOv8 performs well in precise and real-time recognition of objects, making it extremely efficient for detecting regions of interest such as tomato leaves, while ResNet-50’s deep architecture boosts disease classification accuracy by effectively distinguishing between various tomato diseases based on extracted features. Together, they form a powerful framework for accurate tomato disease prediction.
Keywords: Classification, Disease Yolo, GPT3.5, Leaf Extraction, Prediction, Resnet, Tomato