Fig. 5.

Classification between plant and non-plant (background) regions of pre-segmented image using pre-trained color classification models. a Original VIS image of mid-stage arabidopsis shoot in top-view. b Pre-segmented VIS image. c Clustering of (b) into 10 k-means color regions visualized with pseudo-colors. d, e Plant regions computed as median (d) or fusion (e) of eight binary classification results. f Plant regions predicted by eight color classification models. Plots show assignment of region labels to one of two binary color region categories: plant (1) or non-plant (0). The dotted-line indicates the threshold (tsh = 0.5) for assignment of fuzzy label estimates ($FE\in [0,1]$) of some classifiers (glm, gpr, linmod, svmreg) to one of two binary categories, i.e. if $FE\ge 0.5$ then plant (1), if $FE<0.5$ then non-plant (0)