Confidence-Aware Multi-Model Image Classification for Early Disease Detection in Plants

Digital agriculture is essential for enhancing crop yields by integrating modern digital methods to prevent and manage crop diseases. To address this, a deep learning-based Confidence-Aware Multi-Model Image Classification (CAMIC) framework has been developed. CAMIC incorporates FD-Net (Foliar Disease Network) to enable early detection and identification of various plant foliar diseases. Performance testing on the public PlantVillage dataset demonstrated that CAMIC can achieve a high accuracy of up to 97.91%, outperforming existing transfer learning models like ResNet, Inception, Xception, MobileNet, and EfficientNet. This solution has also been implemented as an Android application following the client-server model paradigm.