This project introduces YOLO11n-RP, which is an improved YOLO11n model to make it more compact to achieve good training performance and applies the model to Jetson Nano to run the product recognition and classification system. The model is designed with significantly reduced parameters and weights, thereby minimizing GPU memory consumption during training, compared to the original model. This is a great advantage when deployed on devices with limited resources. With its compact size, YOLO11n-RP is suitable for deployment on edge devices such as Jetson Nano, helping the system operate stably in hardware-limited environments. Although the initial accuracy has decreased due to size optimization, the combination of the argument augmentation method has significantly improved the recognition performance of the model. Specifically, when deploying YOLO11n-RP with argument augmentation on Jetson Nano, the product classification model achieves superior accuracy compared to the original model on the same device. These points confirm the potential of YOLO11n-RP combined with augmentation arguments as an optimal solution, both lightweight and efficient, suitable for real-time applications in resource-constrained environments.