Comparison of the YOLO11n model on embedded platforms Orange Pi5 and Raspberry Pi4B

Context and relevance. Modern trends in technology development are moving towards the transfer of computing from cloud platforms to embedded devices. In this regard, there is a growing need to determine the optimal stack of hardware and software technologies for solving computer vision problems. Goal. Determine the optimal hardware configuration for real-time computer vision tasks based on the criteria of bandwidth, latency, accuracy, and ease of integration. Methods and materials. The paper provides a direct experimental comparison of two embedded platforms for executing the YOLO11n object detection model: a single-board Orange Pi 5 computer with a built-in Rockchip RK3588S neuroprocessor (6 TOPS, INT8) and a Raspberry Pi 4B bundle with an external Google Coral Edge TPU USB accelerator (4 TOPS, INT8). The YOLO11n model was converted from a single PyTorch source file to the target formats RKNN and TensorFlow Lite (full integer quantization) using the tool chains recommended by the manufacturers. The measurements were performed under identical pretreatment conditions. Results. The Orange Pi 5 achieved a steady frame rate of 43-54 FPS (640×640 resolution) with an end-to-end delay of 19-23 ms; accuracy mAP@0.5:0.95 with INT8 quantization, it remains at ~39%. On a Raspberry Pi 4B with Edge TPU at a resolution of 448×448, the delay was 121.5 ms (~8.2 FPS), while 45% of operations are performed on the CPU, which limits scaling. Conclusions. The Orange Pi's gain in speed reaches 6-7 times at a higher resolution, the NPU demonstrates a significant load margin, and the single-board implementation eliminates the overhead of the external interface. The totality of the results allows us to recommend Orange Pi 5 as the preferred platform for building productive edge solutions with real-time object detection.

1. Купцов, А. Р. Квантование как метод оптимизации нейронных сетей // Международный журнал информационных технологий и энергоэффективности. – 2026. – Т. 11, № 4 (66). – С. 195–200.
   Kuptsov, A. R. (2026). Kvantovanie kak metod optimizacii nejronnyh setej [Quantization as a method of neural network optimization]. Mezhdunarodnyj Zhurnal Informacionnyh Tehnologij i Energoeffektivnosti, 11(4), 195–200.
2. Небаба, С. Г. Сверточные нейронные сети семейства YOLO для мобильных систем компьютерного зрения / С. Г. Небаба, Н. Г. Марков // Компьютерная оптика. – 2024. – Т. 16, № 3. – С. 615–631. – DOI: 10.20537/2076-7633-2024-16-3-615-631.
   Nebaba, S. G., & Markov, N. G. (2024). Svertochnye nejronnye seti semejstva YOLO dlya mobil’nyh sistem komp’yuternogo zreniya [Convolutional neural networks of the YOLO family for mobile computer vision systems]. Komp'yuternaya Optika, 16(3), 615–631. - DOI: 10.20537/2076-7633-2024-16-3-615-631
3. Boroumand, A., Ghose, S., Akin, B., et al. (2021). Google neural network models for edge devices: Analyzing and mitigating machine learning inference bottlenecks. arXiv..
4. COCO Dataset. (n.d.). Common Objects in Context. Retrieved March 24, 2026, from https://cocodataset.org/.
5. Coral AI. Edge TPU Models Introduction [Электронный ресурс]. – URL: https://coral.ai/models/ (дата обращения: 24.03.2026).
6. Geng, Y., et al. (2024). A survey on neural network quantization. Proceedings of the 2024 6th International Conference on Computer Information and Big Data Applications.
7. Girshick, R. (2015). Fast R-CNN. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 1440–1448. https://doi.org/10.1109/ICCV.2015.169.
8. Hosseininoorbin, S., Layeghy, S., Kusy, B., Jurdak, R., & Portmann, M. (2023). Exploring Edge TPU for deep feed-forward neural networks. Internet of Things, *23*, Article 100749. https://doi.org/10.1016/j.iot.2023.100749.
9. Jocher, G. (2024). YOLOv11: Next-generation real-time object detection model. Ultralytics GitHub Repository. Retrieved March 24, 2026, from https://github.com/ultralytics/ultralytics.
10. Lee, J., Lee, J., Kim, H., Jeon, S., Yoon, J., Park, H., Park, M., & Ha, H. (2026). TriGen: NPU architecture for end-to-end acceleration of large language models based on SW-HW co-design. arXiv. Retrieved March 24, 2026, from https://arxiv.org/abs/2602.12962.
11. ONNX. (n.d.). Open Neural Network Exchange. Retrieved March 24, 2026, from https://onnx.ai/.
12. Orange Pi. (2023). Orange Pi 5 user manual. Retrieved March 24, 2026, from http://www.orangepi.org/html/hardWare/computerAndMicrocontrollers/details/Orange-Pi-5.html.
13. Performance Evaluation of the Rockchip Systems-on-Chip Through YOLOv4 Object Detection Model. (2023). 2023 IEEE Ural-Siberian Conference on Biomedical Engineering, Radioelectronics and Information Technology (USBEREIT). https://doi.org/10.1109/USBEREIT58508.2023.10158842.
14. Raspberry Pi Ltd. (n.d.). Raspberry Pi OS. Retrieved March 24, 2026, from https://www.raspberrypi.com/software/.
15. Rockchip Electronics Co., Ltd. (2022). RK3588S datasheet version 1.0. Retrieved March 24, 2026, from https://www.rock-chips.com/a/en/products/RK35_Series/.
16. Rokh, B., Azarpeyvand, A., & Khanteymoori, A. (2023). A comprehensive survey on model quantization for deep neural networks in image classification. ACM Transactions on Intelligent Systems and Technology.
17. Seshadri, K., Akin, B., Laudon, J., Narayanaswami, R., & Yazdanbakhsh, A. (2022). An evaluation of Edge TPU accelerators for convolutional neural networks. arXiv.
18. Ultralytics Documentation. (n.d.). YOLOv8 vs YOLOv5 comparison. Retrieved March 24, 2026, from https://docs.ultralytics.com/ru/compare/yolov8-vs-yolov5/.
19. Yang, C., Zhang, R., Huang, L., et al. (2023). A survey of quantization methods for deep neural networks. Chinese Journal of Engineering, *45*(10), 1613–1629.