Nutrition Deficiency In Paddy Leaf Using Neural Network

Abstract

Rice (Oryza sativa) is a primary food crop worldwide and is vulnerable to several foliar diseases that reduce yield. This manuscript proposes a unified, hybrid approach that combines YOLOv8 for real-time localization of symptomatic regions with a lightweight CNN classifier (transfer‑learned EfficientNetB0/MobileNetV2) for fine‑grained disease classification. The pipeline accepts field images of paddy leaves, performs robust preprocessing and augmentation, uses YOLOv8 to detect candidate lesions, crops and normalizes these regions, and classifies each region into one of four consistent disease classes: Healthy, Leaf Blast, Brown Spot, and Bacterial Leaf Blight. We describe dataset preparation, annotation conventions, model architectures, training strategies, and evaluation metrics (mAP, precision, recall, F1, confusion matrix). The proposed hybrid design preserves YOLOv8’s real‑time detection capability while improving classification accuracy on small lesions and visually similar disease symptoms. We provide a clear experimental protocol, recommended hyperparameters, and templates for figures and tables to ensure reproducible and professional presentation.