Crowd Multi Prediction: Single Network for Crowd Counting, Localization and Anomaly Detection

Coskun M. F., AKAR G.

2023 IEEE International Conference on Consumer Electronics, ICCE 2023, Nevada, United States Of America, 6 - 08 January 2023, vol.2023-January identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 2023-January
  • Doi Number: 10.1109/icce56470.2023.10043501
  • City: Nevada
  • Country: United States Of America
  • Keywords: crowd analysis, crowd counting localization and abnormal event detection, multi-task prediction, synthetic crowd dataset generation with GTA-V game
  • Middle East Technical University Affiliated: Yes

Abstract

In this study, we propose a neural network to solve crowd counting, localization and abnormal event detection problems together. Our proposed model combines P2P-Net with a novel crowd anomaly detection module. The final network has a single backbone and multiple head structure. Synthetic datasets are used for training and evaluation. Results show that our model gets high accuracy in crowd anomaly detection task.