Transfer learning is an efficient approach for using information from large well annotated datasets (such as ImageNet) for CADe problems. “Off-the-shelf” features from deep CNN models can be applied ...