ABSTRACT
The variable spatial scale and crowd distribution in crowd images are the main challenges faced by crowd counting problems in recent years. In order to solve the above problems, a crowd counting method based on a multi-scale adaptive network is proposed in this paper. The first 10 layers of the VGG-16 network are used to extract basic features, and the spatial pyramid pooling layer is introduced to make the network adapt to images of any size. Then, multi-scale features are extracted through hybrid dilated convolution, and contrast features are obtained by comparing with basic features. Finally, the weight and density map are calculated to obtain the number of people according to contrast features. The experimental results on the Shanghai Tech and UCF_CC_50 datasets show that, compared with the previous best method, the MAE of the two parts of the Shanghai Tech dataset in this paper is reduced by 1.1 and 0.1, respectively, the MSE is the same in part A, and the part B is reduced by 0.2. On the UCF_CC_50 dataset, the MAE is reduced by 10.9, and the MSE is reduced by 61.4. It shows that the method proposed in this paper has better accuracy and robustness.
- Simonyan K , Zisserman A . Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. Computer Science, 2014.Google Scholar
- He K , Zhang X , Ren S , Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 37(9):1904-16.Google ScholarDigital Library
- Ronghua Liang, Xiangdong Liu, Xiangyin Ma, High-Density Crowd Counting Method Based on SURF[J]. In Journal of Computer Aided Design and Graphics, 2012, 24(12): 1568-1575.Google Scholar
- Viola P, Jones M J, Snow D. Detecting Pedestrians Using Patterns of Motion and Appearance[J]. In International Journal of Computer Vision, 2005, 63(2): 153-161.Google ScholarDigital Library
- Gall J , Member, IEEE, Hough Forests for Object Detection, Tracking, and Action Recognition[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2011, 33(11):2188-202.Google ScholarDigital Library
- Forsyth, David. Object Detection with Discriminatively Trained Part-Based Models.[J]. Computer, 2014.Google Scholar
- Sheng-Fuu, Lin, Jaw-Yeh, Estimation of Number of People in Crowded Scenes Using Perspective Transformation[J]. IEEE Transactions on Systems Man & Cybernetics Part A, 2001.Google ScholarDigital Library
- Bo W , Nevatia R . Detection of Multiple, Partially Occluded Humans in A Single Image by Bayesian Combination of Edgelet Part Detectors. IEEE, 2005.Google Scholar
- Tao Z , Nevatia R , Bo W . Segmentation and Tracking of Multiple Humans in Crowded Environments[J]. 2007.Google Scholar
- Chan A B , Vasconcelos N . Bayesian Poisson Regression for Crowd Counting. IEEE, 2010.Google Scholar
- Ryan D , Denman S , Fookes C B , Crowd Counting Using Multiple Local Features. IEEE, 2009.Google ScholarDigital Library
- H. Idrees, I. Saleemi, C. Seibert, and M. Shah. Multi-Source Multi-Scale Counting in Extremely Dense Crowd Images. In CVPR, 2013, pp. 2547–2554.Google ScholarDigital Library
- C. Wang, H. Zhang, L. Yang, S. Liu, and X. Cao. Deep People Counting in Extremely Dense Crowds. In ACM MM. ACM, 2015, pp. 1299–1302.Google Scholar
- Kumagai S , Hotta K , Kurita T . Mixture of Counting CNNs: Adaptive Integration of CNNs Specialized to Specific Appearance for Crowd Counting[J]. 2017.Google Scholar
- C. Zhang, H. Li, X. Wang, and X. Yang. Cross-Scene Crowd Counting Via Deep Convolutional Neural Networks. In CVPR, 2015, pp. 833-841.Google ScholarCross Ref
- Y. Zhang, D. Zhou, S. Chen, S. Gao, and Y. Ma. Single-Image Crowd Counting Via Multi-Column Convolutional Neural Network. In CVPR, 2016, pp. 589–597.Google ScholarCross Ref
- D. B. Sam, S. Surya, and R. V. Babu. Switching Convolutional Neural Network for Crowd Counting. In CVPR. IEEE, 2017, pp. 4031–4039.Google Scholar
- X. Liu, J. van de Weijer, and A. D. Bagdanov. Leveraging Unlabeled Data for Crowd Counting by Learning to Rank. In CVPR, 2018, pp. 7661–7669.Google Scholar
- Y. Li, X. Zhang, and D. Chen. Csrnet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes. In CVPR, 2018, pp. 1091–1100.Google ScholarCross Ref
- Z. Shen, Y. Xu, B. Ni, M. Wang, J. Hu, and X. Yang. Crowd Counting Via Adversarial Cross-Scale Consistency Pursuit. In CVPR, 2018, pp. 5245–5254.Google ScholarCross Ref
- ZHAI Qiang, WANG Luyang, YIN Baoqun, Crowd Counting Algorithm Based on Scale Adaptive Convolutional Neural Network[J]. Computer Engineering, 2020, 46(2): 250-254, 261.Google Scholar
- J. Hu, L. Shen, and G. Sun. Squeeze-And-Excitation Networks. In CVPR, 2018, pp. 7132–7141.Google ScholarCross Ref
Index Terms
- Research on Crowd Counting Algorithm Based on Multi-scale Adaptive Network
Recommendations
Multi-scale dilated convolution of convolutional neural network for crowd counting
AbstractGrowing numbers of crowd density estimation methods have been developed in scene monitoring, crowd safety and on-site management scheduling. We proposed a method for density estimation of a single static image based on convolutional neural network ...
Crowd Counting based on Multi-level Multi-scale Feature
AbstractCrowd counting has drawn more and more attention for its significance in reality application. However, it’s still a challenging task because of scale variation in images. In this paper, we propose a model to extract and refine features with ...
Atrous convolutions spatial pyramid network for crowd counting and density estimation
AbstractScale variation because of perspective distortion is still a challenge for crowd analysis. To address this problem, an atrous convolutions spatial pyramid network (ACSPNet) is proposed to perform crowd counts and density maps for both ...
Comments