Comparing Computing Platforms for Deep Learning on a Humanoid Robot

  • Alexander BiddulphEmail author
  • Trent Houliston
  • Alexandre Mendes
  • Stephan K. Chalup
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11307)


The goal of this study is to test two different computing platforms with respect to their suitability for running deep networks as part of a humanoid robot software system. One of the platforms is the CPU-centered Intel® NUC7i7BNH and the other is a NVIDIA® Jetson TX2 system that puts more emphasis on GPU processing. The experiments addressed a number of benchmarking tasks including pedestrian detection using deep neural networks. Some of the results were unexpected but demonstrate that platforms exhibit both advantages and disadvantages when taking computational performance and electrical power requirements of such a system into account.


Deep learning Robot vision GPU computing Low powered devices 


  1. 1.
    Abouelfarag, A.A., Nouh, N.M., ElShenawy, M.: Improving performance of dense linear algebra with multi-core architecture. In: 2017 International Conference on High Performance Computing Simulation (HPCS), pp. 870–874, July 2017.
  2. 2.
    Allgeuer, P., Farazi, H., Schreiber, M., Behnke, S.: Child-sized 3D printed IGUS humanoid open platform. In: IEEE-RAS 15th International Conference on Humanoid Robots (Humanoids) (2015)Google Scholar
  3. 3.
    Bainville, E.: GPU benchmarks (2009).
  4. 4.
    Chalup, S.K., Murch, C.L., Quinlan, M.J.: Machine learning with aibo robots in the four legged league of robocup. IEEE Trans. Syst. Man Cybern. Part C 37(3), 297–310 (2007)CrossRefGoogle Scholar
  5. 5.
    Franklin, D.: NVIDIA Jetson TX2 Delivers Twice the Intelligence to the Edge (2017). Accessed 11 Apr 2018
  6. 6.
    Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press, Cambridge (2016)Google Scholar
  7. 7.
    Gregg, C., Hazelwood, K.: Where is the data? why you cannot debate CPU vs. GPU performance without the answer. In: (IEEE ISPASS) IEEE International Symposium on Performance Analysis of Systems and Software, pp. 134–144, April 2011.
  8. 8.
    Houliston, T., et al.: Nuclear: a loosely coupled software architecture for humanoid robot systems. Front. Robot. AI 3(20) (2016).
  9. 9.
    Houliston, T., Metcalfe, M., Chalup, S.K.: A fast method for adapting lookup tables applied to changes in lighting colour. In: Almeida, L., Ji, J., Steinbauer, G., Luke, S. (eds.) RoboCup 2015. LNCS (LNAI), vol. 9513, pp. 190–201. Springer, Cham (2015). Scholar
  10. 10.
    Howard, A.G., et al.: MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. CoRR abs/1704.04861 (2017).
  11. 11.
    Huang, J., et al.: Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors. CoRR abs/1611.10012 (2016).
  12. 12.
    Intel Corporation: Product Brief: Intel®NUC Kit NUC7i7BNH (2017). Accessed 11 Apr 2018
  13. 13.
    Jabbar, A., Farrawell, L., Fountain, J., Chalup, S.K.: Training deep neural networks for detecting drinking glasses using synthetic images. In: Proceedings of the The 24th International Conference On Neural Information Processing, ICONIP 2017, Guangzhou, China, 14–18 November 2017CrossRefGoogle Scholar
  14. 14.
    Javadi, M., Azar, S.M., Azami, S., Shiry, S., Ghidary, S.S., Baltes, J.: Humanoid robot detection using deep learning: a speed-accuracy tradeoff. In: 21st Annual RoboCup International Symposium (2017, in press)Google Scholar
  15. 15.
    Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. In: Pereira, F., Burges, C.J.C., Bottou, L., Weinberger, K.Q. (eds.) Advances in Neural Information Processing Systems (NIPS 2012), vol. 25, pp. 1097–1105. Curran Associates Inc. (2012)Google Scholar
  16. 16.
    Lee, V.W., et al.: Debunking the 100X GPU vs. CPU myth: an evaluation of throughput computing on CPU and GPU. ACM SIGARCH Comput. Archit. News 38(3), 451–460 (2010)CrossRefGoogle Scholar
  17. 17.
    Levine, S., Pastor, P., Krizhevsky, A., Ibarz, J., Quillen, D.: Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection. Int. J. Robot. Res. 37(4), 421–436(2017). Scholar
  18. 18.
    Lin, T., et al.: Microsoft COCO: Common Objects in Context. CoRR abs/1405.0312 (2014).
  19. 19.
    Liu, W., et al.: SSD: Single Shot MultiBox Detector. CoRR abs/1512.02325 (2015).
  20. 20.
    Metcalfe, M., Annable, B., Olejniczak, M., Chalup, S.K.: A study on detecting three-dimensional balls using boosted classifiers. In: Interactive Entertainment 2016, at the Australasian Computer Science Week (ACSW 2016), Canberra, Australia. ACM Digital Library, 2–5 February 2016.
  21. 21.
    Pierson, H.A., Gashler, M.S.: Deep learning in robotics: a review of recent research. Adv. Robot. 31(16), 821–835 (2017). Scholar
  22. 22.
    Speck, D., Barros, P., Weber, C., Wermter, S.: Ball localization for robocup soccer using convolutional neural networks. In: Behnke, S., Sheh, R., Sarıel, S., Lee, D.D. (eds.) RoboCup 2016. LNCS (LNAI), vol. 9776, pp. 19–30. Springer, Cham (2017). Scholar
  23. 23.
    Vanhoucke, V., Senior, A., Mao, M.Z.: Improving the speed of neural networks on CPUs. In: Deep Learning and Unsupervised Feature Learning Workshop, NIPS 2011 (2011)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  • Alexander Biddulph
    • 1
    Email author
  • Trent Houliston
    • 1
  • Alexandre Mendes
    • 1
  • Stephan K. Chalup
    • 1
  1. 1.School of Electrical Engineering and ComputingThe University of NewcastleCallaghanAustralia

Personalised recommendations