Laser & Optoelectronics Progress, Volume. 59, Issue 24, 2415002(2022)
Traffic Sign Detection Algorithm Based on Modified Anchor-Free Model
Figures & Tables(9)
Fig. 5. Frequency distribution of traffic signs with frequency greater than 100 in TT100K
Table 1. Comparison results of detection performance for different sizes of traffic signs on TT100K dataset
View tableTable 1. Comparison results of detection performance for different sizes of traffic signs on TT100K dataset
Method All /% Small /% Medium /% Large /% Params /106 Model size /Mbit FLOPs /109 Faster R-CNN［ 5 ］R 86 77 91 92 134 518 1267 A 82 73 88 86 YOLOv3［ 23 ］R 92 87 95 91 62 237 802 A 85 80 89 82 YOLOv4［ 24 ］R 92 87 95 92 64 214 728 A 88 81 92 86 Zhu et al［ 22 ］R 91 87 94 88 81 418 A 88 82 91 91 CenterNet-Res101［ 11 ］R 91 85 95 94 53 214 768 A 89 78 94 92 CenterNet-Hourglass［ 11 ］R 94 89 96 94 191 765 4672 A 90 83 95 93 Proposed method R 93 88 96 95 16 64 320 A 90 83 95 93
Table 2. Results of ablation experiments of various components on TT100K dataset
View tableTable 2. Results of ablation experiments of various components on TT100K dataset
Model R /% A /% AP /% Params /106 Model Size /Mbit FLOPs /109 Base 89.3 87.2 89.8 15.4 62.2 295.4 Base+RADM 90.8 88.6 91.9 15.7 63.0 296.9 Base+MFFN 92.1 89.4 92.7 15.8 63.6 318.4 Base+RADM+MFFN 92.5 90.3 93.6 16.1 64.4 320.1
Tools
Get Citation
Copy Citation Text
Lü Wei, Zhiyin Liang, Jinghui Chu. Traffic Sign Detection Algorithm Based on Modified Anchor-Free Model[J]. Laser & Optoelectronics Progress, 2022, 59(24): 2415002
Paper Information
Category: Machine Vision
Received: Sep. 1, 2021
Accepted: Oct. 27, 2021
Published Online: Jan. 11, 2023
The Author Email: Jinghui Chu (cjh@tju.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20