Object Detection Based on Fast/Faster RCNN Employing Fully Convolutional Architectures

Table 4

The results of Fast RCNN with skip-layer connections.

Network	Training time	Model size	mAP (%)
GoogleNet	6 s	129 MB	64.0
Inception_v2	11 s	145 MB	65.2
Inception_v3	16 s	191 MB	68.9
ResNet_50	14 s	200 MB	70.8