Geodesic
Re-evaluation of the CNN-based state-of-the-art crowd-counting methods with enhancements
Computer Science and Information Systems, Tome 19 (2022) no. 3.

Voir la notice de l'article provenant de la source Computer Science and Information Systems website

Crowd counting has a range of applications and it is an important task that can help with the accident prevention such as crowd crushes and stampedes in political protests, concerts, sports, and other social events. Many crown counting approaches have been proposed in the recent years. In this paper we compare five deep-learning-based approaches to crowd counting, reevaluate them and present a novel CSRNet-based approach. We base our implementation on five convolutional neural network (CNN) architectures: CSRNet, Bayesian Crowd Counting, DM- Count, SFA-Net, and SGA-Net and present a novel approach by upgrading CSRNet with application of a Bayesian crowd counting loss function and pixel modeling. The models are trained and evaluated on three widely used crowd image datasets, ShanghaiTech part A, part B, and UCF-QNRF. The results show that models based on SFA-Net and DM-Count outperform state-of-the-art when trained and evaluated on the similar data, and the proposed extended model outperforms the base model with the same backbone when trained and evaluated on the significantly different data, suggesting improved robustness levels.
Keywords: Crowd counting, convolutional neural networks, deep learning 
@article{CSIS_2022_19_3_a4,
     author = {Matija Ter\v{s}ek and Ma\v{s}a Kljun and Peter Peer and \v{Z}iga Emer\v{s}i\v{c}},
     title = {Re-evaluation of the {CNN-based} state-of-the-art crowd-counting methods with enhancements},
     journal = {Computer Science and Information Systems},
     publisher = {mathdoc},
     volume = {19},
     number = {3},
     year = {2022},
     url = {http://geodesic.mathdoc.fr/item/CSIS_2022_19_3_a4/}
}
TY  - JOUR
AU  - Matija Teršek
AU  - Maša Kljun
AU  - Peter Peer
AU  - Žiga Emeršič
TI  - Re-evaluation of the CNN-based state-of-the-art crowd-counting methods with enhancements
JO  - Computer Science and Information Systems
PY  - 2022
VL  - 19
IS  - 3
PB  - mathdoc
UR  - http://geodesic.mathdoc.fr/item/CSIS_2022_19_3_a4/
ID  - CSIS_2022_19_3_a4
ER  - 
%0 Journal Article
%A Matija Teršek
%A Maša Kljun
%A Peter Peer
%A Žiga Emeršič
%T Re-evaluation of the CNN-based state-of-the-art crowd-counting methods with enhancements
%J Computer Science and Information Systems
%D 2022
%V 19
%N 3
%I mathdoc
%U http://geodesic.mathdoc.fr/item/CSIS_2022_19_3_a4/
%F CSIS_2022_19_3_a4
Matija Teršek; Maša Kljun; Peter Peer; Žiga Emeršič. Re-evaluation of the CNN-based state-of-the-art crowd-counting methods with enhancements. Computer Science and Information Systems, Tome 19 (2022) no. 3. http://geodesic.mathdoc.fr/item/CSIS_2022_19_3_a4/