Early identification of crop diseases is critical in minimizing the yield losses and have sustainable food production. Traditional methods, like manual scouting and lab testing, tend to be labor intensive, subjective and not applicable to large scale supervising. This paper suggests a hybrid pipeline, which comprises Unmanned Aerial Vehicles (UAVs) plus some sophisticated artificial intelligence (AI) algorithms to identify crop diseases in the field in real-time. The methodology includes the UAV-based data collection, image pre-processing, vegetation index calculation (NDVI and ExG) and hybrid deep learning models of the lesions detection and classification. The trained models were also trained to be used on edge devices to provide real time inference when flying the UAV. The obtained experimental results showed an accuracy in detection of high accuracy (mAP 0.5 = 0.87, F1 = 0.88) and reached 28 FPS on a simple embedded GPU. Explainability tools, e.g. Grad-CAM overlays, verified consistency with expert labels, and geo-referenced severity maps provided accuracy intervention at the field level. The suggested system is capable of effectively closing this gap between laboratory research and practical application in agriculture with the solution being scalable to precision agriculture.