Traffic Sign Detection Based on Improved Faster R-CNN Model

Fig. 10. P-R curves. (a) Improved basic network; (b) improved basic network+multi-scale feature fusion; (c) improved basic network+multi-scale feature fusion+improved anchor generation

Download full size

Fig. 11. Example of detection effect of proposed algorithm. (a) Complex background detection; (b) detection of small targets

Download full size

Table 1. Basic network parameters

View table

Table 1. Basic network parameters

Stage	Output size	Network structure
Conv1	1024×1024	Convolution kernel size: 7×7, 64, stride: 2;max pooling size: 3×3, 64, stride: 2
Conv2	512×512	$[\begin{array}{l} 1 \times 1,128 \\ 1 \times 1,3 \times 3,5 \times 5,7 \times 7,128,1, C = 4 \times 8 \\ 1 \times 1,256 \end{array}] \times 3$
Conv3	256×256	$[\begin{array}{l} 1 \times 1,256 \\ 1 \times 1,3 \times 3,5 \times 5,7 \times 7,256,1, C = 4 \times 8 \\ 1 \times 1,512 \end{array}] \times 3$
Conv4	128×128	$[\begin{array}{l} 1 \times 1,512 \\ 1 \times 1,3 \times 3,5 \times 5,7 \times 7,512,1, C = 4 \times 8 \\ 1 \times 1,1024 \end{array}] \times 3$
Conv5	64×64	$[\begin{array}{l} 1 \times 1,1024 \\ 1 \times 1,3 \times 3,5 \times 5,7 \times 7,1024,1, C = 4 \times 8 \\ 1 \times 1,2048 \end{array}] \times 3$

Table 2. Hardware configuration parameters
View table
Table 2. Hardware configuration parameters
Name Configuration
CPU Intel Core i7 8700K
GPU NVIDIA RTX 2080Ti
RAM/G 32
Hard disk /TB 2

Table 3. Performance comparison of detection algorithms
View table
Table 3. Performance comparison of detection algorithms
Algorithm AP /% Time /s
Traditional Faster R-CNN 80.28 0.33
Ref.[19] 81.13 0.31
Ref.[8] 91.67 5.81
Proposed algorithm 90.83 2.87

Tools

Get Citation

Copy Citation Text

Yi Zhang, Zhiyuan Gong, Wenwen Wei. Traffic Sign Detection Based on Improved Faster R-CNN Model[J]. Laser & Optoelectronics Progress, 2020, 57(18): 181015

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites