Ensemble-based Deep Learning Framework for Robust Sugarcane Leaf Disease Identification

Early and accurate identification of sugarcane leaf diseases is essential to prevent yield loss and ensure sustainable crop management. Manual inspection methods are often inconsistent and impractical for large-scale farming environments. This paper presents a compact and high-accuracy framework for sugarcane leaf disease detection using an ensemble of CNN, ResNet-50, and EfficientNet models. Preprocessed leaf images are used to capture diverse field conditions. Deep features extracted from each architecture are combined through a weighted ensemble strategy to improve classification reliability and reduce individual model bias. Experimental evaluation demonstrates strong classification performance across multiple disease categories. The proposed hybrid approach attains superior performance in terms of accuracy, precision, recall, and F1-score as compared to standalone models, while remaining robust to variations in lighting, background, and leaf orientation. The suggested framework offers a scalable and effective solution for precision agriculture and supports intelligent cropmonitoring systems.