Table 2.

Parameters used for training four different CNN-based algorithms for CAD.

Algorithm	Mask R-CNN	RetinaNet	eDenseYOLO	ASPP U-Net
Stages	Two stages	Single stage	Single stage	Single stage
Backbone	Resnet101	ResNet101	DensNet201	ResNet50
Optimizer	SGD	SGD	Adam	Adam
Learning rate	1e-6	1e-6	1e-3	1e-3
Weight decay	0.0	0.0	0.0005	–
Momentum	0.9	0.9	0.9	0.9, 0.999