SSD及其不同主网络实现的性能差异比较

SSD-Caffe(Official) – VGG16

System	VOC2007 test mAP	FPS (Titan X)	Number of Boxes	Input resolution
Faster R-CNN (VGG16)	73.2	7	~6000	~1000 x 600
YOLO (customized)	63.4	45	98	448 x 448
SSD300* (VGG16)	77.2	46	8732	300 x 300
SSD512* (VGG16)	79.8	19	24564	512 x 512

SSD-TensorFlow – VGG16|

Model	Training data	Testing data	mAP	FPS
SSD-300 VGG-based	VOC07+12 trainval	VOC07 test	0.778	-
SSD-300 VGG-based	VOC07+12+COCO trainval	VOC07 test	0.817	-
SSD-512 VGG-based	VOC07+12+COCO trainval	VOC07 test	0.837	-

SSD - Caffe - MobileNet

Model	mAP	FPS	Remarks
MobileNet-SSD	72.7(VOC0712)	Remains	MS-COCO pretraining

Squeeze-SSD

Model	mAP	FPS	Remarks
SqueezeDet	Remains	-	-
SqueezeNet-SSD	0.643	100(GPU)	-