Crop detection and disease management are being revolutionized in the modern era of precision agriculture by artificial intelligence (AI) and machine learning (ML) techniques. This study proposes a novel deep learning framework that combines ensemble techniques, transfer learning, and conventional neural networks (CNNs) to reliably identify and categorize crops from high-resolution satellite and drone imagery. To assist farmers in making well-informed decisions, our method utilizes various datasets and explainable AI tools like Grad CAM and SHAP to produce clear and easy-to-understand predictions. Tests on Indian agricultural datasets demonstrate enhanced detection robustness and accuracy under various field conditions, such as changing lighting, crop growth stages, and weed density. The proposed system aims to reduce manual intervention, optimize resource allocation, and increase crop yield.