Crowd counting á la Bourdieu, Automated estimation of the number of people

Karolina Przybylek1 and Illia Shkroba2

  1. University of Warsaw, Warsaw
    00-721 Podchorazych 20, Poland
  2. Polish-Japanese Academy of Computer Technology, Warsaw
    02-008 Koszykowa 86, Poland


In recent years, sociologists have taught us how important and emergent the problem of crowd counting is. They have recognised a variety of reasons for this fact, including: public safety (e.g. crushing between people, trampling underfoot, risk of spreading infectious disease, aggression), politics (e.g. police and governent tend to underestimate the number of people, whilst protest organisers tend to overestimate it) and journalism (e.g. accuracy of the estimation of the ground truth supporting an article). The aim of this paper is to investigate models for crowd counting that are inspired by the observations of famous sociologist Pierre Bourdieu. We show that despite the simplicity of the models, we can achieve competitive result. This makes them suitable for low computational power and energy efficient architectures.

Key words

crowd counting, deep learning, mall dataset, habitus

Digital Object Identifier (DOI)

Publication information

Volume 17, Issue 3 (October 2020)
Year of Publication: 2020
ISSN: 1820-0214 (Print) 2406-1018 (Online)
Publisher: ComSIS Consortium

Full text

DownloadAvailable in PDF
Portable Document Format

How to cite

Przybylek, K., Shkroba, I.: Crowd counting á la Bourdieu, Automated estimation of the number of people. Computer Science and Information Systems, Vol. 17, No. 3, 959–982. (2020),