Learning Weight of Losses in Multi-Scale Crowd Counting

Derya Uysal*, Ulug Bayazit

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this work, we improve the state-of-the-art in crowd counting by further developing a recently proposed multi-scale, and multi-task crowd counting approach. While most of the studies treat density-based architectures, this study proposed a point-based method for crowd analysis. We propose automatic, and optimal weight assignment to constituents of the loss function. This approach, which is applied to each patch, ensures that the weight parameters are updated in each epoch, and added to the optimizer with model parameters rather than remaining constant. For validation of our proposed approach, we use three popular crowd counting datasets, ShanghaiTech A, ShanghaiTech B, and UCF_CC_50. The performance of our approach exceeds the performances of the other studies on the ShanghaiTech dataset, and is highly competitive with the performances of the other studies on the UCF CC 50 dataset.

Original languageEnglish
Title of host publication17th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2023 - Proceedings
EditorsTulay Yildirim, Richard Chbeir, Ladjel Bellatreche, Costin Badica
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798350338904
DOIs
Publication statusPublished - 2023
Event17th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2023 - Hammamet, Tunisia
Duration: 20 Sept 202323 Sept 2023

Publication series

Name17th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2023 - Proceedings

Conference

Conference17th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2023
Country/TerritoryTunisia
CityHammamet
Period20/09/2323/09/23

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Keywords

  • Automatic Weighted Loss
  • Crowd Counting
  • Multi Scale
  • Point Supervision

Fingerprint

Dive into the research topics of 'Learning Weight of Losses in Multi-Scale Crowd Counting'. Together they form a unique fingerprint.

Cite this