Table 1.

Comparison of multi-object detection methods based on deep learning.

Method	Device	Model	Region Proposal	Test (Region Proposal + Detection)
R-CNN	CPU & GPU	Mulit-ConvNet Multi-Classifier	Selective Search	2 s + 47 s
SPP-net	CPU & GPU	Mulit-ConvNet SPP Pooling	Selective Search	2 s + 2.3 s
Fast R-CNN	CPU & GPU	Share ConvNet RoI Pooling	Selective Search	2 s + 0.32 s
Faster R-CNN	GPU	End-to-End Share ConvNet RoI Pooling	Region Proposal Network	0.01 s+ 0.2 s