Skip to main content
Fig. 4 | Plant Methods

Fig. 4

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. 4

Regression of total root length (mm) per image as derived from manually human labelled masks and as predicted by U-Net models (Table 4) on the mixed test dataset. a, c and e are U-Net models with default (UNetGNRes), SE-ResNeXt-101 and EfficientNet-b6 decoders trained without augmented data, respectively; b, d and f are the corresponding models trained with augmented data (+ aug). Formulas indicate the slope and offset of linear regressions; shaded areas represent 95% confidence interval. Models predict less root length than manually labelled masks. The 1:1 line is shown as a dashed line; R.2 values indicate goodness of fit (n = 69)

Back to article page

Plant Methods

ISSN: 1746-4811