Skip to main content
. 2019 Dec 10;9:18738. doi: 10.1038/s41598-019-55373-7

Figure 2.

Figure 2

Architecture of eDenseYOLO with a backend network of DenseNet201. The output layers of eDenseYOLO, i.e., YOLO v2 with DenseNet201, were modified for improved robustness with respect to disease-pattern size. If the input is 256 × 256, the feature map for the last layer is 8 × 8, 16 × 16, or 32 × 32 with a skip connection.