Dense pedestrian detection algorithm based on improved Yolov5-DCN

Yu-ning Wei; Qi Hu

doi:10.1117/12.2685500

28 July 2023 Dense pedestrian detection algorithm based on improved Yolov5-DCN

Yu-ning Wei, Qi Hu

Proceedings Volume 12716, Third International Conference on Digital Signal and Computer Communications (DSCC 2023); 127160P (2023) https://doi.org/10.1117/12.2685500
Event: Third International Conference on Digital Signal and Computer Communications (DSCC 2023), 2023, Xi'an, China

Abstract

Paper investigates an enhanced approach on YOLOv5 method that solves the challenges posed by complex environmental backgrounds, intensive repeated target detection, and varying pedestrian sizes. In the backbone network, feature extraction mainly depends on the texture information and shape of the target. We use deformable convolutional network (DCN) to replace the traditional convolution.;In Non-Maximum Suppression(NMS),The Generalized Intersection over Union(GIOU) used by original network is in place of through Distance Intersection over Union Loss(DIOU) solve problem of the densely populated repeated test.We use crowdhuman training data set to evaluate the effectiveness of the algorithm.We observed that the detection accuracy of the improved Yolov5-DCN model was 83.8%, which was 1.6% higher than that of the basic model. Moreover, it can effectively improve the accuracy of pedestrian target detection in dense scenes, especially for the detection of dense occluded targets, and the effect is significantly improved.

Citation Download Citation

Yu-ning Wei and Qi Hu "Dense pedestrian detection algorithm based on improved Yolov5-DCN", Proc. SPIE 12716, Third International Conference on Digital Signal and Computer Communications (DSCC 2023), 127160P (28 July 2023); https://doi.org/10.1117/12.2685500

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available