Figure 6. Comparison of the model performance to four independent human observers.

(A) shows number of positive cells detected by model (red) or counted by observers in the validation images. (B) shows the relative differences of the model to the mean of the observers’ manual counts. The data in (B) is summarized in boxplots in (C) and (D). In (A) and (B) the x-axis represents images, which are sorted in order from smallest to largest number of mean counted cells. The dashed line in (B), (C) and (D) represents the mean counted cells over all images and all observers; the red squares in (C) and (D) show the mean. White circles in (B) represent counts of a person on a respective image, whereas the filled red circles represent the measurements of a model. In (D) red dots represent counts of cells in validation images.