A Deep Learning-Based Web Platform for Multi-Crop Disease Detection and Treatment Recommendation

Abstract

Both bacterial and viral pathogens cause plant diseases that are a serious threat to agricultural production and food security. The traditional method of diagnosing diseases is based on physical inspection, which takes a great deal of time and is complicated by lack of resources. Though various deep learning techniques have shown results, these innovations require large computational resources, low level of explainability, and operate almost exclusively offline. Crop Care AI addresses the problems represented by traditional as well as deep learning technologies by implementing the use of EfficientNet-B0 for effective classification of plant diseases. The system uses the advanced localization algorithm based on color identification for detecting infected areas without sophisticated description requirements. The online web application is safe and it provides access to multiple users that can check their prediction histories and see where the disease has spread in the most effective way. Testing shows that the system has accuracy rates of 93.44%, which proves its efficiency in diagnosis making.