Abstract: The rising threat of rice diseases such as blast, brown spot, sheath blight, and bacterial leaf blight has emphasized the need for early, accurate, and scalable detection methods. This research brings together insights from 30 peer-reviewed studies to provide a consolidated view of how computational intelligence is being applied in rice disease diagnosis. Recent advancements highlight the use of deep learning models—including CNNs, ResNet, DenseNet, EfficientNet, Vision Transformers, and object detection frameworks like YOLO—alongside traditional machine learning techniques such as Support Vector Machines (SVM) and Random Forests (RF). In certain cases, hybrid systems (e.g., CNN-SVM) have demonstrated almost flawless classification performance, while advanced architectures like DenseNet201 and Vision Transformers have reported accuracy levels exceeding 98%. Other studies focus on practical aspects such as IoT-enabled monitoring, mobile-based deployment, and image segmentation for precise localization of disease-affected regions.

Although these developments are highly encouraging, challenges remain in transitioning from controlled research environments to real-world applications. Key issues include generalizing across diverse field conditions, maintaining computational efficiency on resource-limited devices, and ensuring models perform reliably across different geographic and climatic variations. This paper provides a critical evaluation of available methodologies, summarizes key achievements, and discusses limitations that researchers and practitioners must address. Finally, it outlines potential future directions for developing more adaptive, resource-efficient, and farmer-friendly rice disease detection systems that can support sustainable agricultural practices.

Keywords: Rice disease detection, CNN, precision agriculture, machine learning, IoT in farming.

Downloads: | DOI: 10.17148/IARJSET.2025.12830