Skip to main content
Fig. 3 | Plant Methods

Fig. 3

From: Semantic segmentation of plant roots from RGB (mini-) rhizotron images—generalisation potential and false positives of established methods and advanced deep-learning models

Fig. 3

Receiver operating characteristic (ROC) curves of true vs false positive rates (FPR) on MR images of a training data set of Zea mays roots (“ATTRACT 1” project), without augmentation, b the mixed validation data set with augmentation (+ aug), and c the unseen Cichorium intybus (RootPainter) data set (Additional file 2). A diagonal dashed line indicates the dummy classifier, values above the line are better, values below are worse than a random classifier. The “elbow” on the left indicates a more “conservative” classifier, such as Adaptive Thresholding, while being on the right indicates a more “liberal” classifier, such as Frangi Vesselness. The closer the “elbow” is to the upper left corner (0,1), the better the model. ROCs of all methods are shown in different colours; SegRoot + aug is not shown for clarity (it largely overlaps with the dummy classifier)

Back to article page

Plant Methods

ISSN: 1746-4811