Advertisement

Deep Learning in Vehicle Pose Recognition on Two-Dimensional Images

  • Dmitry Yudin
  • Ekaterina Kapustina
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 875)

Abstract

The paper describes usage of deep neural network architectures such as VGG, ResNet and InceptionV3 for the classification of small images. Each image may contain one of four vehicle pose categories or background. An iterative procedure for training a neural network is proposed, which allows us to quickly tune the network using wrongly classified images on test sample. A dataset of more than 23,000 marked images was prepared, of which 70% of images were used as a training sample, 30% as a test sample. On the test sample, the trained deep convolutional neural networks are ensured the recognition accuracy for all classes of at least 93.9%, the classification precision for different vehicle poses and background was from 85.29% to 100.0%, the recall was from 81.9% to 100.0%. The computing experiment was carried out on a graphics processor using NVIDIA CUDA technology. It showed that the average processing time of one image varies from 3.5 ms to 15.9 ms for different architectures. Obtained results can be used in software for image recognition of road conditions for unmanned vehicles and driver assistance systems.

Keywords

Image recognition Classification Vehicle pose Deep learning Convolutional neural network 

Notes

Acknowledgment

This article is written in the course of the grant of the President of the Russian Federation for state support of young Russian scientists № MK-3130.2017.9 (contract № 14.Z56.17.3130-MK) on the theme “Recognition of road conditions on images using deep learning”.

References

  1. 1.
    Hou, T., Wang, S., Qin, H.: Vehicle matching and recognition under large variations of pose and illumination. In: International Conference Computer Vision and Pattern Recognition, pp. 24–29 (2009)Google Scholar
  2. 2.
    Prokaj, J., Medioni, G.: 3-D model based vehicle recognition. In: Workshop on Applications of Computer Vision (WACV), pp. 1–7 (2009)Google Scholar
  3. 3.
    Jayawardena, S., Hutter, M., Brewer N.: A novel illumination-invariant loss for monocular 3D pose estimation. In: International Conference on Digital Image Computing: Techniques and Applications, pp. 37–44 (2011)Google Scholar
  4. 4.
    Glasner, D., Galun, M., Alpert, S., Basri, R., Shakhnarovich G.: Viewpoint-aware object detection and pose estimation. In: IEEE International Conference on Computer Vision, pp. 1275–1282 (2011)Google Scholar
  5. 5.
    Penate-Sanchez, A., Moreno-Noguer, F., Andrade-Cetto, J., Fleuret, F.: LETHA: learning from high quality inputs for 3D pose estimation in low quality images. In: IEEE Second International Conference on 3D Vision, pp. 517–524 (2014)Google Scholar
  6. 6.
    Novotny, D., Larlus, D., Vedaldi, A.: Learning 3D object categories by looking around them. In: IEEE International Conference on Computer Vision, pp. 5228–5237 (2017)Google Scholar
  7. 7.
    Sedaghat, N., Brox, T.: Unsupervised generation of a viewpoint annotated car dataset from videos. In: IEEE International Conference on Computer Vision, pp. 1314–1322 (2015)Google Scholar
  8. 8.
    Ozuysal, M., Lepetit, V., Fua P.: Pose estimation for category specific multiview object localization. In: IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2009, pp. 778–785 (2009)Google Scholar
  9. 9.
    Teney, D., Piater, J.: Continuous pose estimation in 2D images at instance and category levels. In: International Conference on Computer and Robot Vision, pp. 121–127 (2013)Google Scholar
  10. 10.
    Pedersoli, M., Tuytelaars, T.: A scalable 3D HOG model for fast object detection and viewpoint estimation. In: IEEE Second International Conference on 3D Vision, pp. 163–170 (2014)Google Scholar
  11. 11.
    Bakry, A., Elgaaly, T., Elhoseiny, M., Elgammal, A.: Joint object recognition and pose estimation using a nonlinear view-invariant latent generative model. In: IEEE Winter Conference on Applications of Computer Vision (WACV), pp. 1–9 (2016)Google Scholar
  12. 12.
    Gu, H.-Z., Lee, S.-Y.: Car model recognition by utilizing symmetric property to overcome severe pose variation. Mach. Vis. Appl. 24, 255–274 (2013)CrossRefGoogle Scholar
  13. 13.
    Rubino, C., Crocco, M., Del Bue, A.: 3D object localisation from multi-view image detections. IEEE Trans. Pattern Anal. Mach. Intell. 40(6), 1281–1294 (2018)Google Scholar
  14. 14.
    Li, W., Luo, Y., Wang, P., Qin, Z., Zhou, H., Qiao, H.: Recent advances on application of deep learning for recovering object pose. In: Proceedings of the 2016 IEEE International Conference on Robotics and Biomimetics, Qingdao, China, pp. 1273–1280 (2016)Google Scholar
  15. 15.
    Sochor, J., Herout, A., Havel, J.: BoxCars: 3D boxes as CNN input for improved fine-grained vehicle recognition. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3006–3015 (2016)Google Scholar
  16. 16.
    Xue, Y., Qian, X.: Vehicle detection and pose estimation by probabilistic representation. In: ICIP 2017, pp. 3355–3359 (2017)Google Scholar
  17. 17.
    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna Z.: Rethinking the inception architecture for computer vision. In: ECCV, arXiv:1512.00567 (2016)
  18. 18.
    Kaiming, H., Xiangyu, Z., Shaoqing, R., Jian S.: Deep residual learning for image recognition. In: ECCV, arXiv:1512.03385 (2015)
  19. 19.
    Simonyan, K., Zisserman, A.: Very deep convolutions for large-scale image recognition. In: ICLR, arXiv:1409.1556 (2015)
  20. 20.
    Yudin, D.A., Kapustina, E.O.: Dataset containing four car pose categories and background (2017). https://yadi.sk/d/xjQKIoyU3NNVyt. Last Accessed 11 May 2018
  21. 21.
    Olson, D.L., Delen D.: Advanced Data Mining Techniques, 1st edn. Springer, Heidelberg (2008)Google Scholar
  22. 22.
    Yudin, D., Knysh, A.: Vehicle recognition and its trajectory registration on the image sequence using deep convolutional neural network. In: The International Conference on Information and Digital Technologies, pp. 435–441 (2017)Google Scholar
  23. 23.
    DeepClassificationTool. Deep image classification tool based on Keras. https://github.com/yuddim/deepClassificationTool. Last Accessed 11 May 2018

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  1. 1.Belgorod State Technological University named after V.G. ShukhovBelgorodRussia

Personalised recommendations