EfficientNet-B4 Based Deep Learning for Automated Paddy Leaf Disease Classification

Paddy leaf diseases can have a major impact on the yield and quality of paddy crops, which emphasizes the need of early and precise disease detection for efficient disease management. In this paper, we present an EfficientNet architecture-based deep learning method for paddy leaf disease classification. Compound scaling of EfficientNet makes it possible to preserve computational economy while achieving better performance. We assess our methods using a dataset that includes four illness classifications: Brown Spot, Hispa, Leaf Blast, and Healthy. Our suggested approach outperforms both previous deep learning models and orthodox machine learning techniques, with an accuracy of 85.7%. EfficientNet performs better than other CNN models in terms of accuracy, precision, recall, and F1-score, according to comparative research. Because of its remarkable accuracy, efficiency, and resilience, EfficientNet performs better than other solutions, making it a viable option for automatic paddy leaf disease classification in real-world agricultural situations.