RDHCNet – Residual Depthwise Hybrid Convolutional Network for Robust Crop Disease Diagnosis

Abstract

Crop diseases are a major danger to the world's food security because they reduce crop productivity and farmer revenue. Early detection and preventative measures can reduce these losses. This study suggests CropNet Hybrid, a deep learning model that can identify 38 crop disease classes in a variety of plant species after being trained on the PlantVillage dataset. In contrast to earlier research that only looks at classification, our system incorporates a carefully chosen knowledge-based prevention module, giving farmers practical advice. On test data, the model, which was implemented as a hybrid CNN architecture with depthwise separable convolutions and residual blocks, achieved an accuracy of more than 93.27%. The framework is a useful tool for smart agriculture since it is implemented as a FastAPI microservice and offers real-time detection and prevention guidance.