Figure 6.
(A,F) An image example and two image regions in it. (B,G) The corresponding ground truth. (C–E) The outputs of CNN, GCN, and VTG-Net. (H) CNN-only outputs by our CNN (cyan arrows highlight some vessels with multiple class assignments, and orange arrows mark limited vessel connectivity in the CNN-only outputs, which are rectified by our GCN as shown in I).