1. A-CCNN: ADAPTIVE CCNN FOR DENSITY ESTIMATION AND CROWD COUNTING
Saeed Amirgholipour1, Xiangjian He1; Wenjing Jia1, Dadong Wang2, Michelle Zeibots1
1Global Big Data Technologies Centre, University of Technology Sydney, Australia
2 Quantitative Imaging, CSIRO Data61, Australia

2. Objectives Improve the accuracy of CCNN method
Easily handle large-scale variations in people’s sizes when appearing in images;
Facilitate to generate local density maps within a crowd scene.

3. What's the problem? CCNN counting: Upper part: 2 persons Lower part:
A-CCNN counting:
9 persons
3 persons

4. What's the problem? CCNN counting: Upper part: 7 persons Lower part:
A-CCNN counting:
14 persons
3 persons

5. Literature Review on Crowd Counting
CNN-based approaches
Network Property
Basic
Scale aware models
context aware model
multi task model
Training Process
Patch-based
Whole image based

6. Literature Review on Crowd Counting
Methods
Category
Network
property
Inference
process
Idrees [11]
Basic
Patch-based
Zhang et al. [17]
Multi-task
Onoro et al. [2]
Scale-aware
Shang et al. [6]
Context-aware
Whole image-based
Sindagi et al. [5]
Sam et al. [16]

7. Progress to Date – Adaptive CCNN

8. Tiny Face detection
[8] P. Hu and D. Ramanan, “Finding tiny faces,” in Proceedings of the CVPR. IEEE, 2017, pp. 1522–1530.

9. Fazzy Model

10. Adaptive CCNN - UCSD Dataset
Comparison of the MAE results results between A-CCNN and state-of-the-art crowd counting on UCSD crowd-counting dataset [10]
Methods
Maximal
Downscale
Upscale
Minimal
Avg
Density Learning [13]
1.70
1.28
1.59
2.02
1.64
Count Forest [14]
1.43
1.30
1.62
1.49
Arteta et al. [15]
1.24
1.31
1.69
Zhang et al. [16]
1.26
1.52
Switch-CNN [5] -
CCNN [2]
1.79
1.13
1.50
1.51
A-CCNN
1.36
1.04
1.48
1.35

11. Adaptive CCNN - UCF Dataset
Comparison of the MAE results between A-CCNN and state-of-the-art crowd counting on UCF CC dataset [11].
Methods
MAE
Density learning [13]
493.4
Idrees et al. [11]
419.5
Zhang et al. [16]
467.0
MCNN [17]
377.6
Hydra-CCNN [2]
333.73
Switch-CNN [5]
318.1
CCNN [2]
488.67
A-CCNN
367.3

12. Adaptive CCNN - STF Dataset

13. Adaptive CCNN - STF Dataset
Comparison of the MAE results results between A-CCNN and state-ofthe-art crowd counting on STF [12].
Methods C5 C9
Farhood et al. [12]
2.28
2.67
CCNN [2]
3.90
4.23
A-CCNN
1.69
1.87

14. Some Results

15. Some Results

16. Some Results

17. Some Results

18. Thanks

19. Some References
Wu, Shuang, Hang Su, Hua Yang, Shibao Zheng, Yawen Fan, and Qin Zhou. "Bilinear dynamics for crowd video analysis." Journal of Visual Communication and Image Representation 48 (2017):
Kumagai, Shohei, Kazuhiro Hotta, and Takio Kurita. "Mixture of Counting CNNs: Adaptive Integration of CNNs Specialized to Specific Appearance for Crowd Counting." arXiv preprint arXiv: (2017).
Sindagi, Vishwanath A., and Vishal M. Patel. "A survey of recent advances in cnn-based single image crowd counting and density estimation." Pattern Recognition Letters 107 (2018): 3-16.

20. Some References
Cassol, Vinícius J., Soraia R. Musse, Cláudio R. Jung, and Norman I. Badler. "Crowd Analysis Based on Computer Vision." In Simulating Crowds in Egress Scenarios, pp Springer, Cham, 2017.
Shao, Jing, Chen Change Loy, Kai Kang, and Xiaogang Wang. "Crowded scene understanding by deeply learned volumetric slices." IEEE Transactions on Circuits and Systems for Video Technology 27, no. 3 (2017):
Rodriguez, Mikel, Josef Sivic, and Ivan Laptev. "The Analysis of High Density Crowds in Videos." In Group and Crowd Behavior for Computer Vision, pp
Wang, Xiaogang, and Chen-Change Loy. "Deep Learning for Scene-Independent Crowd Analysis." In Group and Crowd Behavior for Computer Vision, pp
Zeng, Lingke, Xiangmin Xu, Bolun Cai, Suo Qiu, and Tong Zhang. "Multi-scale convolutional neural networks for crowd counting." In Image Processing (ICIP), 2017 IEEE International Conference on, pp IEEE, 2017.

21. back
[11] H. Idrees, I. Saleemi, C. Seibert, and M. Shah, Multi-source multi-scale counting in extremely dense crowd images," in Proceedings of the CVPR, 2013, pp

22. back
[17] Y. Zhang, D. Zhou, S. Chen, S. Gao, and Y. Ma,"Single-image crowd counting via multi-column convolutional neural network," in Proceedings of the CVPR, 2016, pp

23. back
[2] D. Onoro-Rubio and R.J. Lpez-Sastre, "Towards perspective-free object counting with deep learning," in Proceedings of the ECCV. Springer, 2016, pp

24. back
[6] C. Shang, H. Ai, and B. Bai,"End-to-end crowd counting via joint learning local and global count," in Proceedings of the ICIP. IEEE, 2016, pp

25. back
[16] V.A. Sindagi and V.M. Patel, "Cnn-based cascaded multi-task learning of high-level prior and density estimation for crowd counting," in Advanced Video and Signal Based Surveillance (AVSS). IEEE, 2017, vol. 14, pp. 1-6.

26. back
[5] D.B. Sam, S. Surya, and R.V. Babu,"Switching convolutional neural network
for crowd counting," in CVPR, 2017, vol. 1/3, p. 6