Advertisement

BlurNet: Keeping Collected Data Private with a Neural Network Based Pipeline

Conference paper
  • 371 Downloads
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1196)

Abstract

Data collection process is required to keep personal information of third parties private. This mandatory obligation has been recently enforced in a growing number of countries by appropriate laws, which forbid recording certain types of data. On the other hand, machine learning solutions based on neural networks, which are used in automotive industry, require a vast amount of data to learn from. The article addresses the problem of collecting camera frames in the form of a video file, where registered pedestrian faces and vehicle license plates are visible to a degree. This allows one to associate them with a particular person, hence they contain personal data. The pipeline for object detection and blurring algorithm is proposed. Processing images is being done by proposed BlurNet neural network based on YOLOv3 architecture. The article describes adaptations for a given data format, as well as modifications to improve the accuracy. Blurring algorithm is described as well. In order to maintain sufficient level of privacy, conducted experiments provide numerical answer regarding the performance of such solution.

Keywords

Object detection Neural network Faces License plates Blurring Image processing Data collection Automotive Privacy 

Notes

Acknowledgment

Research is partially funded by Polish Ministry of Science and Higher Education (MNiSW) Project No. 0014/DW/2018/02 and carried out in cooperation of Aptiv Services Poland S.A. – Technical Center Kraków and AGH University of Science and Technology – Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering.

References

  1. 1.
  2. 2.
    Xu, L., Jiang, C., Wang, J., Yuan, J., Ren, Y.: Information security in big data: privacy and data mining. IEEE Access 2, 1149–1176 (2014).  https://doi.org/10.1109/ACCESS.2014.2362522CrossRefGoogle Scholar
  3. 3.
    Newton, E., Sweeney, L., Malin, B.: Preserving privacy by de-identifying face images. IEEE Trans. Knowl. Data Eng. 17, 232–243 (2005).  https://doi.org/10.1109/TKDE.2005.32CrossRefGoogle Scholar
  4. 4.
    Muraki, T., Oishi, S., Ichino, M., Echizen, I., Yoshiura, H.: Anonymizing face images by using similarity-based metric. In: 2013 International Conference on Availability, Reliability and Security, pp. 517–524 (2013).  https://doi.org/10.1109/ARES.2013.68
  5. 5.
    Ren, Z., Lee, Y.J., Ryoo, M.S.: Learning to anonymize faces for privacy preserving action detection. CoRR abs/1803.11556 (2018). http://arxiv.org/abs/1803.11556
  6. 6.
    Wu, Y., Yang, F., Xu, Y., Ling, H.: Privacy-protective-GAN for privacy preserving face de-identification. J. Comput. Sci. Technol. 34, 47–60 (2019).  https://doi.org/10.1007/s11390-019-1898-8CrossRefGoogle Scholar
  7. 7.
    Girshick, R.B., Donahue, J., Darrell, T., Malik, J.: Rich feature hierarchies for accurate object detection and semantic segmentation. CoRR abs/1311.2524 (2013). http://arxiv.org/abs/1311.2524
  8. 8.
    Girshick, R.B.: Fast R-CNN. CoRR abs/1504.08083 (2015). http://arxiv.org/abs/1504.08083
  9. 9.
    Ren, S., He, K., Girshick, R., Sun, J.: Faster R-CNN: towards real-time object detection with region proposal networks, pp. 91–99 (2015). http://papers.nips.cc/paper/5638-faster-r-cnn-towards-real-time-object-detection-with-region-proposal-networks.pdf
  10. 10.
    Lin, T., Goyal, P., Girshick, R.B., He, K., Dollár, P.: Focal loss for dense object detection. CoRR abs/1708.02002 (2017). http://arxiv.org/abs/1708.02002
  11. 11.
    Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S.E., Fu, C., Berg, A.C.: SSD: single shot multibox detector. CoRR abs/1512.02325 (2015). http://arxiv.org/abs/1512.02325
  12. 12.
    Redmon, J., Divvala, S.K., Girshick, R.B., Farhadi, A.: You only look once: unified, real-time object detection. CoRR abs/1506.02640 (2015). http://arxiv.org/abs/1506.02640
  13. 13.
    Redmon, J., Farhadi, A.: Yolo9000: better, faster, stronger. arXiv preprint arXiv:1612.08242 (2016)
  14. 14.
    Redmon, J., Farhadi, A.: Yolov3: an incremental improvement. arXiv (2018)Google Scholar
  15. 15.
    Kingma, D., Ba, J.: Adam: a method for stochastic optimization. In: International Conference on Learning Representations (2014)Google Scholar
  16. 16.
    Geiger, A., Lenz, P., Stiller, C., Urtasun, R.: Vision meets robotics: the KITTI dataset. Int. J. Robot. Res. (IJRR) 32, 1231–1237 (2013)CrossRefGoogle Scholar
  17. 17.
    Powers, D.: Evaluation: from precision, recall and f-measure to ROC, informedness, markedness & correlation. J. Mach. Learn. Technol. 2, 2229–3981 (2011).  https://doi.org/10.9735/2229-3981CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.AGH University of Science and TechnologyKrakówPoland
  2. 2.APTIV Services Poland S.A.KrakówPoland

Personalised recommendations