Pedestrian Target Detection in Subway Scene Using Improved YOLOv5s Algorithm

Xiuzai Zhang; Ye Qiu; Chen Zhang

doi:10.3788/LOP213000

Laser & Optoelectronics Progress, Volume. 60, Issue 6, 0610013(2023)

Pedestrian Target Detection in Subway Scene Using Improved YOLOv5s Algorithm

Xiuzai Zhang^1,2、*, Ye Qiu¹, and Chen Zhang¹

¹School of Electronic and Information Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu, China

²Jiangsu Province Atmospheric Environment and Equipment Technology Collaborative Innovation Center, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. YOLOv5s algorithm model

Download full size

Fig. 2. Schematic diagram of YOLOv5s algorithm detection

Download full size

Fig. 3. Improved algorithm YOLOv5s model

Download full size

Fig. 4. D-SE attention module

Download full size

Fig. 5. Structure diagram of D-SE attention module

Download full size

Fig. 6. Structure diagram of depth residual shrinkage network

Download full size

Fig. 7. Structure diagram of S-ASPP

Download full size

Fig. 8. Comparison of Validation-loss

Download full size

Fig. 9. Pedestrian detection effects of different models. (a)-(c) Original image; (d)-(f) YOLOv3 detection effect; (g)-(i) YOLOv4 detection effect; (j)-(l) YOLOv5s detection effect; (m)-(o) improved YOLOv5s detection effect

Download full size

Table 1. Performance comparison of different modules

View table

Table 1. Performance comparison of different modules

Method	Backbone	Neck	NMS	$f_{A P}$ /%	FPS /（frame·s^-1）
1	CSPDarkNet	SPP+PAN	DIOU-NMS	87.6	35.6
2	CSPDarkNet	SPP+PAN	SDIOU-NMS	88.2	33.9
3	DRSN-CSPDarkNet	SPP+PAN	DIOU-NMS	88.6	33.1
4	DRSN-CSPDarkNet	SPP+PAN	SDIOU-NMS	89.1	31.1
5	CSPDarkNet	S-ASPP+PAN	DIOU-NMS	87.8	35.1
6	CSPDarkNet	S-ASPP+PAN	SDIOU-NMS	88.7	33.4
7	DRSN-CSPDarkNet	S-ASPP+PAN	DIOU-NMS	88.6	33.8
8	DRSN-CSPDarkNet	S-ASPP+PAN	SDIOU-NMS	89.6	30.4

Table 2. Comparison of detection indexes of different models
View table
Table 2. Comparison of detection indexes of different models
Method P /% R /% $f_{A P}$ /% FPS /（frame·s^-1）
Improved YOLOv5s 89.7 88.9 89.6 30.4
YOLOv5s 89.1 87.2 87.6 35.6
YOLOv4 86.4 86.1 85.6 32.9
YOLOv3 85.8 83.2 81.2 26.4
YOLOv3-Tiny 84.1 81.1 83.1 37.4
R-CNN 80.1 82.1 79.8 20.4

Tools

Get Citation

Copy Citation Text

Xiuzai Zhang, Ye Qiu, Chen Zhang. Pedestrian Target Detection in Subway Scene Using Improved YOLOv5s Algorithm[J]. Laser & Optoelectronics Progress, 2023, 60(6): 0610013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites