Pedestrian Shoes Detection Algorithm Based on SSD

Table 1. Number of objects and images in different datasets
View table
Table 1. Number of objects and images in different datasets
Name Training dataset Test dataset
Object Image Object Image
Shoe 2223 1134 247 127

Table 2. Detection effect of different feature layers
View table
Table 2. Detection effect of different feature layers
Conv4_3 Conv7 Conv8_2 Conv9_2 Conv10_2 Conv11_2 AP FPS
√ √ √ √ √ √ 0.831 31
√ √ √ √ √ 0.834 31
√ √ √ √ 0.829 32
√ √ √ 0.837 33
√ √ 0.830 35
√ 0.811 38

Table 3. Detection effect of different detection networks
View table
Table 3. Detection effect of different detection networks
Model AP FPS
SSD300-VGG 0.831 25
SSD512-VGG 0.863 11
FSSD-VGG 0.866 20
RFB-VGG 0.862 11
SSD300-V 0.891 33

Table 4. Detection effect of different extraction networks
View table
Table 4. Detection effect of different extraction networks
Model AP FPS
SSD-VGG 0.831 25
SSD-EfficientNet_b3 0.862 51
SSD-MobileNet_V2 0.774 91
SSD300-V 0.891 33

Tools

Get Citation

Copy Citation Text

Geng Pengzhi, Yang Zhixiong, Zhang Jiajun, Tang Yunqi. Pedestrian Shoes Detection Algorithm Based on SSD[J]. Laser & Optoelectronics Progress, 2021, 58(6): 610009

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites