Figure 4 | Scientific Reports

Figure 4

From: Reproducibility of abnormality detection on chest radiographs using convolutional neural network in paired radiographs obtained within a short-term interval

Figure 4

Architecture of the eDenseYOLO of which backend network is DenseNet201. The output layers of eDenseYOLO, which is You Only Look Once v2 with DenseNet201, were modified for improved robustness to the variable size of disease patterns. If the input resolution was 256 × 256, the feature map for the last layer was 8 × 8, 16 × 16, and 32 × 32 with skip connection.

Back to article page