Measuring Crowd Collectiveness

Bolei Zhou1, Xiaoou Tang1, and Xiaogang Wang2

1Department of Informaiton Engineering, 2Department of Electronic Engineering, The Chinese University of Hong Kong.


[PDF(CVPR)] [PDF(TPAMI)] [Presentation] [Demo Video] [Dataset] [Code] [gKLT tracker]


1. Introduction

As shown in Figure 1A, crowds in nature have a variety of shapes, dynamics and scales. To quantitatively analyze the crowd, we face two challenges:

Figure 1. A) Collective motions of bacterial colony, fish shoal, bird flock, sheep herd, athletic group, and traffic flow. B) One common spatially coherent structure, i.e., Collective Manifolds, emerging in these different crowds. Since individuals in a crowd system only coordinate their behaviors in their neighborhood, individuals at a distance may have low velocity correlation even though they are on the same collective manifold (such as the red individual and the green individual).

This work aims at proposing a general measurement of collectiveness for different crowd systems and its efficient computation by characterizing the structural properties of the collective manifold in various crowds. Collectiveness is defined as the degree of individuals acting as a union in collective motion. Under this measurement, crowd dynamics could be accurately estimated and quantified into different dynamics categories as shown in Figure 2.

Figure 2. Crowd videos are organized into three dynamics categories based on their estimated collectiveness value.

Based on the proposed collectiveness, we further propose a algorithm called Collective Merging to detect collective motions from clutters. As shown in Figure 3, applications of the proposed collectiveness and algorithm include:

Figure 3. There are wide applications for the proposed collectiveness descriptor.


2. Demo Video

Here is the video to demonstrate the applications of the proposed collectiveness. You could also download this video at here.


3. Reference

If you use our codes (source codes or gKLT tracker) or dataset, please cite our paper.

Bolei Zhou, Xiaoou Tang, and Xiaogang Wang. "Measuring Crowd Collectiveness." Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2013, oral paper) 

Bolei Zhou, Xiaoou Tang, Hepeng Zhang, and Xiaogang Wang. "Measuring Crowd Collectiveness." The IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI, regular paper) 


Last update: Jan 15, 2014