License plate recognition (LPR) systems face challenges in accuratly detecting and recognizing plates from specific regions, particularly with limited real-world datasets. This study proposes a novel approach for Moroccan license plate recognition using the lightweight YOLOv10n model and synthetic data augmentation. We present a two stage method: initial training on the AOLP dataset for general license plate detection, followed by fine-tuning with synthetically generated Moroccan plate images. Advanced augmentation techniques were employed to enhance the model’s robustness and generalization capabilities. Our system achieves a remarkable balance between speed and accuracy, demonstrating a mean Average Precision (mAP) of 99.5% on a diverse test set of Moroccan license plates. This research contributes to the field of LPR by showcasing the effectiveness of combining state-of-the-art object detection models with synthetic data in overcoming dataset limitations and achieving high performance for region-specific applications.