Skip to main content

Table 4 Algorithmic leaf counting results obtained using the method in [15]

From: Citizen crowds and experts: observer variability in image-based plant phenotyping

  Algorithm versus annotator Algorithm versus annotator Annotator versus reference
Training error Testing error Inter-observer error
DiC ↓ 0.00 (1.07) −  0.04 (1.31) 0.21 (0.75)
|DiC| ↓ 0.61 (0.88) 0.88 (0.96) 0.46 (0.62)
MSE ↓ 1.163 1.700 0.600
R2 0.933 0.895 0.964
  1. Four metrics are reported. We first compare between the algorithm and the 728 images in the training set (ie. how well the algorithm learns). Then we compare how well the algorithm predicts counts on a testing set of 130 images (also used in this study) comparing the algorithm with the counts of the annotator (that also was involved in deriving annotations for the training set). Lastly we compare the annotator (the data of which we used to train the algorithm and was not involved in this study) with the reference observer used throughout in this study