A Systematic Review of Standalone and Hybrid Convolutional Neural Network Models for Rice Leaf Disease Detection

Rice leaf diseases pose a serious threat to global food security, reducing yield quality and quantity across major rice-producing regions. Conventional detection methods relying on manual inspection are often time-consuming, subjective, and inefficient. In recent years, deep learning particularly Convolutional Neural Networks (CNNs) has emerged as a transformative tool for automated plant disease diagnosis. This study presents a systematic review of recent advancements in standalone and hybrid CNN–machine learning (CNN–ML) models for rice leaf disease detection. Following the PRISMA 2020 framework, four major databases (IEEE Xplore, Scopus, Web of Science, and Google Scholar) were searched for studies published between 2020 and 2025. A total of 37 peer-reviewed articles were included after applying rigorous inclusion and exclusion criteria. Data were extracted on model architectures, datasets, performance metrics, and computational efficiency, and were analyzed both quantitatively and qualitatively using descriptive statistics and comparative synthesis. Results reveal that standalone CNN models such as ResNet, DenseNet, and EfficientNet achieved a median accuracy of 96.4%, while hybrid CNN–ML models (e.g., CNN–SVM and CNN–Random Forest) recorded a median accuracy of 95.2% with notably reduced inference times and model sizes. Lightweight models like MobileNet and EfficientNet-Lite demonstrated optimal trade-offs between accuracy and resource efficiency, supporting their suitability for edge deployment in low-resource environments. Despite these advancements, only 30% of studies utilized field-based datasets, highlighting a persistent generalization gap between controlled and real-world conditions. Furthermore, explainable AI tools were employed in merely one-third of studies, limiting interpretability and trust in AI-driven diagnostics. This review emphasized the growing maturity of deep learning for smart agriculture while identifying critical gaps in dataset diversity, interpretability, and deployment feasibility. It concludes that future research should prioritize standardized datasets, explainable and lightweight architectures, and energy-efficient edge intelligence to ensure sustainable, inclusive, and transparent AI adoption in rice disease management.