International Journal of Science, Engineering and TechnologyAuthors: Roobannidhi K A, Revin Kumar R, Sakthi GaneshB
Abstract: Agriculture plays a vital role in the Indian economy, with sugarcane being one of the major commercial crops. Sugarcane is affected by diseases such as Mosaic, Red Rot, Rust, and Yellow Leaf Disease, which significantly reduce crop yield and quality. Early detection is essential for effective crop management, yet traditional manual inspection methods are time-consuming and inaccessible to rural farmers who lack expert guidance. This paper presents CaneCare, a deep learning-based system for automatic detection and classification of sugarcane leaf diseases from images. A curated dataset of 2,521 images across five classes—Healthy, Mosaic, RedRot, Rust, and Yellow—was used to train and evaluate three convolutional neural network architectures: EfficientNet-B0, ResNet-50, and MobileNetV3, all fine-tuned using transfer learning. ResNet-50 achieved the highest test accuracy of 94.71% (F1 = 0.947), while MobileNetV3 achieved 94.18% accuracy with only 1.5M parameters, making it the preferred choice for mobile deployment. EfficientNet-B0 provided balanced performance at 90.34%. A complete end-to-end deployment pipeline was implemented comprising a FastAPI backend hosted on the Render cloud platform and a React Native mobile application (CaneCare) built with Expo. Grad-CAM visualisations confirm that model predictions are grounded in biologically meaningful, disease-specific leaf features. The system provides real-time disease predictions with confidence scores and treatment recommendations, demonstrating a practical, accessible, and scalable pathway for AI-driven precision agriculture in resource-constrained environments.
