Skip to main content
. 2019 May 7;9:7024. doi: 10.1038/s41598-019-43171-0

Figure 2.

Figure 2

Flowchart of our deep learning-based method. The method consists of four convolutional layers (1~4) for extracting feature maps, two convolutional layers (5~6), a RPN containing k2(C + 1) ROIs to detect position-sensitive score maps and make final results, as well a softmax layer to output classifications and localizations.