AgriXplain: Explainable AI Crop Intelligence for Smarter, Sustainable Farming

Early detection of agricultural diseases is important for reducing crop damage, particularly in rural or remote regions. This research work provides an explainable Convolutional Neural Network (CNN) technique for agricultural disease alert system that uses Sentinel-2 multispectral data, NDVI and Red Edge NDVI (NDRE) fusion to produce continuous notifications. It increases spectral learning features like stressed plant signals into the raw red, near-infrared and red edge bands. It also increases the capacity to recognize healthy and unhealthy crops. Grad-CAM's module help farmers and agronomists to evaluate model results by recognizing the primary spectral data related to each prediction increases transparency and trust. A client-side storage system allows monitoring the fields and disease analysis queries offline, that can be updated with the server when a connection is available. The evaluations using merged sentinel-2 data show that combining NDVI and NDRE increases the accuracy of detection and produces feature activation maps makes its efficient for early stress detection. This technology was created for adaptive communication with agricultural systems and it provides accurate crop health signals using satellite images.