%0 Conference Paper
%A Abubaker, Mohammed
%A Alsadder, Zubayda
%A Abdelhaq, Hamed
%A Chraibi, Mohcine
%A Boltes, Maik
%A Alia, Ahmed
%T A Novel Dataset for Detecting Pedestrian Heads in Crowds Using Deep Learning Algorithms
%M FZJ-2024-07217
%D 2024
%X The automatic detection of pedestrian heads in crowded environments is crucial for various crowd analysis and management tasks, including crowd counting, density estimation, pedestrian trajectory extraction, and behavior detection. Despite advancements in deep learning algorithms for object detection, existing studies struggle with pedestrian head detection in crowded situations such as railway platforms and event entrances, where risks frequently arise. One main reason for the poor head detection performance is the underrepresentation of such scenarios in existing datasets. These scenarios are particularly challenging due to variations in lighting conditions, viewpoints, occlusions, scale changes, indoor/outdoor environments,and weather conditions.To narrow this gap, we introduce a novel, diverse, and high-resolution dataset of human heads in crowds at Railway Platforms and Event Entrances, named the RPEE-Heads dataset.
%B Traffic and Granular Flow
%C 2 Dec 2024 - 5 Dec 2024, Lyon (France)
Y2 2 Dec 2024 - 5 Dec 2024
M2 Lyon, France
%F PUB:(DE-HGF)6
%9 Conference Presentation
%R 10.34734/FZJ-2024-07217
%U https://juser.fz-juelich.de/record/1034451