Figure 9.

Fully convolutional networks (FCNs) for the semantic segmentation of DFUs (50). The model learns features with forward and backward learning for segmentation. C1–C8 are convolutional layers and P1–P5 are max-pooling layers.