Journal of the Korea Society of Computer and Information (한국컴퓨터정보학회논문지)

Korean Society of Computer Information (한국컴퓨터정보학회)
권호 표지
DOI QR코드

DOI QR Code

Improving Performance of YOLO Network Using Multi-layer Overlapped Windows for Detecting Correct Position of Small Dense Objects

  • Yu, Jae-Hyoung (School of Electronic Engineering, Soongsil University) ;
  • Han, Youngjoon (Department of Smart Systems Software, Soongsil University) ;
  • Hahn, Hernsoo (School of Electronic Engineering, Soongsil University)
  • Received : 2019.01.25
  • Accepted : 2019.02.22
  • Published : 2019.03.29
Download PDF
⟨ Previous Next ⟩

Abstract

This paper proposes a new method using multi-layer overlapped windows to improve the performance of YOLO network which is vulnerable to detect small dense objects. In particular, the proposed method uses the YOLO Network based on the multi-layer overlapped windows to track small dense vehicles that approach from long distances. The method improves the detection performance for location and size of small vehicles. It allows crossing area of two multi-layer overlapped windows to track moving vehicles from a long distance to a short distance. And the YOLO network is optimized so that GPU computation time due to multi-layer overlapped windows should be reduced. The superiority of the proposed algorithm has been proved through various experiments using captured images from road surveillance cameras.

Keywords

CPTSCQ_2019_v24n3_19_f0001.png 이미지

Fig. 2. Process of YOLO network model

CPTSCQ_2019_v24n3_19_f0002.png 이미지

Fig. 3. YOLO Network Structure

CPTSCQ_2019_v24n3_19_f0003.png 이미지

Fig. 4. Concept for multi-layer overlapped windows

CPTSCQ_2019_v24n3_19_f0004.png 이미지

Fig. 5. Sample CCTV image captured on the road

CPTSCQ_2019_v24n3_19_f0005.png 이미지

Fig. 6. Detection Result in Small Dense Vehicles case

CPTSCQ_2019_v24n3_19_f0006.png 이미지

Fig. 7. Example of assign and keep ID each vehicle on entrance and exit image

CPTSCQ_2019_v24n3_19_f0007.png 이미지

Fig. 8. Sub-Window and Intersection Area

CPTSCQ_2019_v24n3_19_f0008.png 이미지

Fig. 9. Setup method for identical vehicle decision

CPTSCQ_2019_v24n3_19_f0009.png 이미지

Fig. 10. Vehicle Identity Decision inside Intersection Area

CPTSCQ_2019_v24n3_19_f0010.png 이미지

Fig. 11. Setup Multi-Windows in Experiment

CPTSCQ_2019_v24n3_19_f0011.png 이미지

Fig. 12. Detection result for small dense vehicle

CPTSCQ_2019_v24n3_19_f0012.png 이미지

Fig. 13. Voting detection result according to distance

CPTSCQ_2019_v24n3_19_f0013.png 이미지

Fig. 14. Comparison for vehicle tracking result using own ID from entrance to way out

CPTSCQ_2019_v24n3_19_f0014.png 이미지

Fig. 15. Relationship accurate and FPS for changing filters

CPTSCQ_2019_v24n3_19_f0015.png 이미지

Fig. 1. (a) Comparison mAP and time of YOLO v2 with other algorithms [12] (b) YOLO v3 [17]]

Table. 1. Vehicle Detection Result

CPTSCQ_2019_v24n3_19_t0001.png 이미지

Table. 2. Vehicle Tracking Result

CPTSCQ_2019_v24n3_19_t0002.png 이미지

Table. 3. Comparison accurate and time according to change filters

CPTSCQ_2019_v24n3_19_t0003.png 이미지

References

  1. S. Russell and P. Norvig, "Artificial Intelligence : A Modern Approach," NJ, USA: Prentice Hall Press, 2009.
  2. O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. C. Berg, and L. Fei-Fei, "ImageNet large scale visual recognition challenge," Int. J. Comput. Vis., vol. 115, no. 3, pp. 211-252, Apr. 2015. https://doi.org/10.1007/s11263-015-0816-y
  3. S. Ioffe and C. Szegedy, "Batch normalization : Accelerating deep network training by reducing internal covariate shift," In Proc. ICML, 2015.
  4. T. Young, D. Hazarika, S. Poria, and E. Cambria, "Recent Trends in Deep Learning Based Natural Language Processing," arXiv preprint arXiv : 1708.02709. 2017.
  5. K. He, G. Gkioxari, P. Dollar, and R. Girshick, "Mask R-CNN," ICCV, pp. 2980-2988. IEEE, 2017.
  6. O. Ronneberger, P. Fischer, and T. Brox, "Unet: Convolutional Networks for Biomedical Image Segmentation," International Conference on Medical image computing and computer-assisted intervention, pp. 234-241, 2015.
  7. V. Badrinarayanan, A. Kendall, and R. Cipolla, "SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation," arXiv preprint arXiv: 1511.00561, 2015.
  8. Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, "Gradientbased learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998. https://doi.org/10.1109/5.726791
  9. K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," In Proc. ICLR, 2015.
  10. A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems, 2012, pp. 1097-1105.
  11. C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich, "Going deeper with convolutions," In Proc. CVPR, 2015.
  12. R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 580-587, 2014.
  13. R. Girshick, "Fast r-cnn," Proceedings of the IEEE International Conference on Computer Vision, 2015, pp. 1440-1448.
  14. W. Liu et al., "SSD: Single shot multibox detector," In Proc. ECCV, pp. 21-37, 2016.
  15. J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once : Unified, real-time object detection," In Proc. CVPR, 2016.
  16. J. Redmon and A. Farhadi, "YOLO9000 : Better, Faster, Stronger", Proceedings of the IEEE conference on computer vision and pattern recognition, pp.779-788, 2016.
  17. J. Redmon and A. Farhadi, "YOLOv3: An Incremental Improvement", arXiv preprint arXiv : 1804.02767, 2018.