Optics and Precision Engineering, Volume. 30, Issue 14, 1764(2022)
Lightweight pedestrian detection for multiple scenes
Figures & Tables(11)
Table 1. Performance comparison of the YOLOX-L and its improved algorithms
View tableView in ArticleTable 1. Performance comparison of the YOLOX-L and its improved algorithms
Algorithm Parameters
/M
Size
/MB
Average precision FPS
（Frame·s-1）
YOLOX-L 47.1 207 69.47% 68 基准模型 53.3 231 72.08% 65 剪枝模型 9.8 47 68.53% 84 微调模型 9.8 47 71.56% 84
Table 2. Performance comparison of the YOLOX-S and its improved algorithms
View tableView in ArticleTable 2. Performance comparison of the YOLOX-S and its improved algorithms
Algorithm Parameters
/M
Size
/MB
Average precision FPS
（Frame·s-1）
YOLOX-S 7.3 35 67.93% 86 基准模型 8.5 39 68.51% 86 剪枝模型 2.6 10.1 67.73% 93 微调模型 2.6 10.1 68.23% 93
Table 3. Performance comparison of different algorithm
View tableView in ArticleTable 3. Performance comparison of different algorithm
Algorithm Backbone Size
/MB
Average precision FPS
（Frame·s-1）
PCN［ 15 ］VGG16 — 66.80% 1.3 Faster-RCNN VGG16 522 70.37% 4 PAMS-FCN VGG16 — 71.44% 27 YOLOv4 Darknet53 246 66.11% 66 YOLOX Darknet53 207 69.47% 68 YOLOv4-Mobilenet Mobilenet 51 64.80% 84 YOLOv4-Ghostnet Ghostnet 43 65.15% 85 YOLOX-tiny Darknet53 20 66.36% 90 本文方法L Darknet53 47 71.56% 84 本文方法S Darknet53 10 68.13% 93
Tools
Get Citation
Copy Citation Text
Yunzuo ZHANG, Wenbo LI, Wei GUO, Zhouchen SONG. Lightweight pedestrian detection for multiple scenes[J]. Optics and Precision Engineering, 2022, 30(14): 1764
Paper Information
Category: Information Sciences
Received: Apr. 19, 2022
Accepted: --
Published Online: Sep. 6, 2022
The Author Email: ZHANG Yunzuo (zyz2016@stdu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20