Rice leaf diseases represent a significant obstacle to global rice production, underscoring the critical need for robust and precise detection methodologies to facilitate prompt interventions. This research meticulously evaluates an extensive range of cutting-edge convolutional neural network (CNN) frameworks, including DenseNet, EfficientNet, ResNet, MobileNet, VGG, Inception, NASNet, Xception, and ConvNeXt, aiming to identify the most effective deep learning model for the accurate classification of common rice leaf diseases such as “leaf blight, brown spot, and leaf smut”. Our comprehensive experimental results indicate that DenseNet201 outperforms all other evaluated models, achieving a high accuracy rate of 99.74%. Building upon this superior model, we introduce RiLeDDS (Rice Leaf Disease Detection System), an advanced system architecture tailored to support practical implementation. RiLeDDS encompasses a wide range of modules, including data acquisition, preprocessing, disease classification, data management, analytical reporting, decision support, and ensuring secure user engagement by protecting data during interaction. By offering accurate and timely decision support through analytical reporting and data visualization, the proposed system holds substantial potential for improving rice disease management, thereby enabling agricultural stakeholders to make well-informed decisions and implement timely interventions.