Fast and Accurate Hand Visual Detection by Using a Spatial-Channel Attention SSD for Hand-Based Space Robot Teleoperation

Table 3

Comparison of object detection methods on VOC 2007 and VOC 2012 test dataset. In this table, the SCA-SSD is compared with some state-of-the-art methods and other SSD-based methods to illustrate the promising performance on object detection.

Method	Backbone	Input size	mAP	Device	FPS
Fast RCNN [21]	ResNet-101		76.4	K40	2.4
YOLO v4 [25]	DarkNet-19		78.2	Titan X	81
R-FCN [38]	ResNet-101		79.5	K40	5.8
StairNet [39]	VGG-16		78.8	Titan X	30
SSD300 [13]	VGG-16		77.2	2080Ti	119
Baseline [37]	VGG-16		77.7	2080Ti	119
DSSD321 [26]	ResNet-101		78.6	Titan X	9.5
FSSD300 [28]	VGG-16		78.8	1080Ti	65.8
FA-SSD300 [40]	ResNet-101		78.3	—	34.7
FF-SSD300 [27]	VGG-16		78.9	—	43
Shift SSD300 [41]	VGG-16		78.3	Titan X	77
ESSD++300 [24]	VGG-16		79.2	—	52
DF-SSD300 [29]	Dense-32-S-1		78.9	Titan X	11.6
RSSD300 [30]	VGG-16		78.5	—	35
SCA-SSD300	VGG-16		79.9	2080Ti	104