A Deep Learning Approach to Vascular Structure Segmentation in Dermoscopy Colour Images

Figure 7

Visualization of segmentation precision over 100 epochs. Colour vascular structures in training patch slices (left), accompanying ground-truth masks (middle), and the predictions (right): (a) 10 epochs, (b) 76 epochs, and (c) 95 epochs.

(a)

(b)

(c)