Optimizing Crowd Counting on Low Compute Device Using Heterogeneous Convolution Filter

Optimizing Crowd Counting on Low Compute Device Using Heterogeneous Convolution Filter
  IJETT-book-cover           
  
© 2022 by IJETT Journal
Volume-70 Issue-3
Year of Publication : 2022
Authors : Christopher Alvin, Gede Putra Kusuma
https://doi.org/10.14445/22315381/IJETT-V70I3P208

How to Cite?

Christopher Alvin, Gede Putra Kusuma, "Optimizing Crowd Counting on Low Compute Device Using Heterogeneous Convolution Filter," International Journal of Engineering Trends and Technology, vol. 70, no. 3, pp. 66-74, 2022. Crossref, https://doi.org/10.14445/22315381/IJETT-V70I3P208

Abstract
The massive growth of population increases potential security risks in the crowded area. To monitor these phenomena, crowd counting can be one of the solutions. But existing crowd counting approach still requires a high compute device. In this work, a mobile device focused density-based crowd counting is proposed. To tackle high resource consumption, the proposed model reduces the number of parameters by using a Heterogeneous Convolution Filter, resulting in lower computation and faster counting time. The experiment is done on three datasets, the ShanghaiTech Part A, ShanghaiTech Part B, UCF_CC_50. Extensive experiments are conducted on mobile devices to have an overview of the performance of the model on mobile devices. The performance of the proposed model compared to CSRNet on mobile are similar on each dataset by only having a bigger MAE by 0.64%, 3%, and 10.56% on UCF_CC_50, ShanghaiTech Part A, and ShanghaiTech Part B, respectively but with better latency by 29.83%, 44.85%, 50%, respectively, and better battery consumption by 25.71%, 44.17% and 50.79% respectively. The proposed model successfully improves the speed of counting the number of people in an image, with a slightly higher MAE compared to the CSRNet on mobile devices.

Keywords
Crowd Counting, Convolutional Neural Network, Heterogeneous Convolution Filter, Deep Learning, Mobile Computing.

Reference
[1] X. Ma, S. Du, and Y. Liu, A Lightweight Neural Network For Crowd Analysis Of Images With Congested Scenes, in 2019 IEEE International Conference on Image Processing (ICIP), (2019) 979–983, doi: 10.1109/ICIP.2019.8803062.
[2] A. Unnikrishnan, A. Narayan, and H. Guruprasad, A Review of the Advances in Population Estimation and Human Behaviour Estimation Using Computer Vision and Mobile Phone Data, Int. J. Comput. Trends Technol., 46(1) (2017) 37–41. doi: 10.14445/22312803/IJCTT-V46P108.
[3] List of human stampedes and crushes, Wikipedia. https://en.wikipedia.org/wiki/List_of_human_stampedes_and_crushes (accessed Oct. 08, 2021).
[4] V. A. Sindagi and V. M. Patel, A survey of recent advances in CNN-based single image crowd counting and density estimation, Pattern Recognit. Lett., 107 (2018) 3–16. doi: 10.1016/j.patrec.2017.07.007.
[5] A. Khan, J. Ali Shah, K. Kadir, W. Albattah, and F. Khan, Crowd Monitoring and Localization Using Deep Convolutional Neural Network: A Review, Appl. Sci., 10(14) (2020). 4781. doi: 10.3390/app10144781.
[6] N. Ilyas, A. Shahzad, and K. Kim, Convolutional-Neural Network-Based Image Crowd Counting: Review, Categorization, Analysis, and Performance Evaluation, Sensors, 20(1) (2019) 43. doi: 10.3390/s20010043.
[7] H. H. Lin and K. T. Win, HPC Feature for Crowd Outdoor Scenes Estimation, Int. J. Comput. Trends Technol.,67(8) (2019) 1–6., doi: 10.14445/22312803/IJCTT-V67I8P101.
[8] C. C. Loy, K. Chen, S. Gong, and T. Xiang, Crowd Counting and Profiling: Methodology and Evaluation, (2013) 347–382.
[9] V. Lempitsky and A. Zisserman, Learning To Count Objects in Images, in Advances in Neural Information Processing Systems, 23 (2013) [Online]. Available: https://proceedings.neurips.cc/paper/2010/file/fe73f687e5bc5280214e0486b273a5f9-Paper.pdf.
[10] S. Ren, K. He, R. Girshick, and J. Sun, Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks, IEEE Trans. Pattern Anal. Mach. Intell., 39(6) (2017)1137–1149. doi: 10.1109/TPAMI.2016.2577031.
[11] S. Basalamah, S. D. Khan, and H. Ullah, Scale Driven Convolutional Neural Network Model for People Counting and Localization in Crowd Scenes, IEEE Access, 7 (2019) 71576–71584.doi: 10.1109/ACCESS.2019.2918650.
[12] D. Babu Sam, S. V. Peri, M. Narayanan Sundararaman, A. Kamath, and V. B. Radhakrishnan, Locate, Size and Count: Accurately Resolving People in Dense Crowds via Detection, IEEE Trans. Pattern Anal. Mach. Intell., (2020) 1–1. doi: 10.1109/TPAMI.2020.2974830.
[13] P. F. Felzenszwalb, R. B. Girshick, D. McAllester, and D. Ramanan, Object Detection with Discriminatively Trained Part-Based Models, IEEE Trans. Pattern Anal. Mach. Intell.,32(9) (2010) 1627–1645.Sep. 2010, doi: 10.1109/TPAMI.2009.167.
[14] A. B. Chan and N. Vasconcelos, Bayesian Poisson regression for crowd counting, in 2009 IEEE 12th International Conference on Computer Vision, (2009) 545–551. doi: 10.1109/ICCV.2009.5459191.
[15] H. Idrees, I. Saleemi, C. Seibert, and M. Shah, Multi-source Multi-scale Counting in Extremely Dense Crowd Images, in 2013 IEEE Conference on Computer Vision and Pattern Recognition, (2013) 2547–2554, doi: 10.1109/CVPR.2013.329.
[16] D. G. Lowe, Object recognition from local scale-invariant features, in Proceedings of the Seventh IEEE International Conference on Computer Vision, 2 (1999) 1150–1157, doi: 10.1109/ICCV.1999.790410.
[17] K. Simonyan and A. Zisserman, Very Deep Convolutional Networks for Large-Scale Image Recognition, (2014). [Online]. Available: http://arxiv.org/abs/1409.1556.
[18] Y. Li, X. Zhang, and D. Chen, CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes, in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2018) 1091–1100, doi: 10.1109/CVPR.2018.00120.
[19] Z. Shi et al., Crowd Counting with Deep Negative Correlation Learning, in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2018) 5382–5390, doi: 10.1109/CVPR.2018.00564.
[20] P. Singh, V. K. Verma, P. Rai, and V. P. Namboodiri, HetConv: Heterogeneous Kernel-Based Convolutions for Deep CNNs, in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2019) 4830–4839, doi: 10.1109/CVPR.2019.00497.
[21] Y. Zhang, D. Zhou, S. Chen, S. Gao, and Y. Ma, Single-Image Crowd Counting via Multi-Column Convolutional Neural Network, in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2016). 589–597. doi: 10.1109/CVPR.2016.70.