notesum.ai

Published at November 27

RPEE-HEADS: A Novel Benchmark for Pedestrian Head Detection in Crowd Videos

cs.CV

cs.LG

Released Date: November 27, 2024

Authors: Mohamad Abubaker¹, Zubayda Alsadder¹, Hamed Abdelhaq¹, Maik Boltes², Ahmed Alia²

Aff.: ¹Department of Information Technology, An-Najah National University, P4110257 Nablus, Palestine; ²Institute for Advanced Simulation, Forschungszentrum Jülich, 52425 Jülich, Germany

Arxiv: http://arxiv.org/abs/2411.18164v1

Model	Initial Learning Rate	Final Learning Rate	Input Size	Iterations	Library^∗
YOLOv7x	0.01	0.1	640	25,000	Official YOLOv7 [18]
YOLOv8x	0.01	0.01	640	25,000	Ultralytics [54]
YOLOv9-E	0.01	0.0001	640	30,000	Official YOLOv9 [20]
RT-DETR	0.01	0.01	640	6,000	Ultralytics
Faster R-CNN	0.003	0.0003	1,333	320,000	Detectron2 [55]
RetinaNet-101	0.01	0.0001	1,333	90,000	Detectron2
Cascade R-CNN	0.02	0.002	1,333	270,000	Detectron2
Fast R-CNN	0.001	0.00001	1,333	100,000	Detectron2
^∗ The library utilized for implementing, training, and evaluating the corresponding model.