notesum.ai

Published at November 26

DGNN-YOLO: Dynamic Graph Neural Networks with YOLO11 for Small Object Detection and Tracking in Traffic Surveillance

cs.CV

cs.LG

Released Date: November 26, 2024

Authors: Shahriar Soudeep¹, M. F. Mridha¹, Md Abrar Jahin², Nilanjan Dey³

Aff.: ¹Department of Computer Science, American International University-Bangladesh, Dhaka, 1229, Bangladesh; ²Physics and Biology Unit, Okinawa Institute of Science and Technology Graduate University (OIST), Okinawa, 904-0412, Japan; ³Department of Computer Science & Engineering, Techno International New Town, New Town, Kolkata, 700156, India

Arxiv: http://arxiv.org/abs/2411.17251v1

Model	Precision	Recall	mAP@0.5	mAP@0.5:0.95	Training time (s)	Parameters (in millions)
Faster R-CNN	0.7582	0.6075	0.7830	0.4976	40342.34	$\approx 18.9$
YOLO5	0.7789	0.5191	0.5917	0.4679	26894.90	$\approx 9.2$
YOLO8	0.7055	0.6085	0.6175	0.4926	21967.85	$\approx 11.2$
YOLO9	0.8354	0.5330	0.6152	0.4903	19504.32	$\approx 7.3$
YOLO10	0.7786	0.5096	0.6045	0.4852	21967.85	$\approx 8.1$
YOLO11	0.8176	0.5248	0.6107	0.4871	17040.80	$\approx 9.5$
Proposed DGNN-YOLO	0.8382	0.6875	0.7830	0.6476	118344.00	$\approx 20.1$