Fig. 8
Original photo, annotation, segmentation output from U-Net and errors. To illustrate the errors the false positives are shown in red and the false negatives are shown in green. This image is a subregion of a larger image for which U-Net got the best (highest) \(F_1\). The segmentation also contains roots which were missed by the annotator. We were able to confirm this by having the annotator review these particular errors