Skip to main content
. 2019 Feb 2;5(2):26. doi: 10.3390/jimaging5020026

Figure 3.

Figure 3

Network predictions on the STARE dataset. The top row shows the image, segmentation masks and ground truth for the image that scored best when STARE was used to train and test the model; the bottom row shows the worst. For comparison, the cross-validation results from training the model with DRIVE and CHASE_DB1 are shown.