Investigated factors are model architecture, tiling method (FS = fixed-stride, OBP = object-based positioning, OBP+OIC = object-based positioning + object integrity constraint), training data set size (coded in the y-axis labels) and prediction threshold (coded in colors). Crosses depict mean, vertical lines within boxes median values. (PDF)</p
<p>The horizontal dashed line indicates the threshold value (4 × 1400 = 5600) for the stopping crite...
<p>Each subfigure corresponds to one case of labeled examples. (A) MA obtained using 10% labeled exa...
Black arrows indicate substantial changes in comparison to Fig 9 (prediction score threshold of 0.9)...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Results for prediction thresholds 0.90 and 0.98 are nearly identical. Blue arrows indicate shift fro...
PT = Prediction Threshold, Dice = Dice’s coefficient, FS = fixed-stride, OBP = object-based position...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Since all metric results take values between 0 and 1, relatively high and stable values of Recall an...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Effects of tiling method, training data set size and prediction threshold on the segmentation perfor...
<p>Boxplot showing the distribution of recall, precision and AUC values for 1000 prediction models g...
Black arrows indicate substantial changes in comparison to Fig 10 (same prediction score but Mask R-...
<p>The black line represents the breakpoints median values (median = 7 in A group; median = 40.5 in ...
<p>For UTGB and WTGB, boxplots were built considering the deviance values from each of the ten cross...
<p>The horizontal dashed line indicates the threshold value (4 × 1400 = 5600) for the stopping crite...
<p>Each subfigure corresponds to one case of labeled examples. (A) MA obtained using 10% labeled exa...
Black arrows indicate substantial changes in comparison to Fig 9 (prediction score threshold of 0.9)...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Results for prediction thresholds 0.90 and 0.98 are nearly identical. Blue arrows indicate shift fro...
PT = Prediction Threshold, Dice = Dice’s coefficient, FS = fixed-stride, OBP = object-based position...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Since all metric results take values between 0 and 1, relatively high and stable values of Recall an...
Blue arrows indicate shift from fixed-stride to object-based tile positioning, green arrows indicate...
Effects of tiling method, training data set size and prediction threshold on the segmentation perfor...
<p>Boxplot showing the distribution of recall, precision and AUC values for 1000 prediction models g...
Black arrows indicate substantial changes in comparison to Fig 10 (same prediction score but Mask R-...
<p>The black line represents the breakpoints median values (median = 7 in A group; median = 40.5 in ...
<p>For UTGB and WTGB, boxplots were built considering the deviance values from each of the ten cross...
<p>The horizontal dashed line indicates the threshold value (4 × 1400 = 5600) for the stopping crite...
<p>Each subfigure corresponds to one case of labeled examples. (A) MA obtained using 10% labeled exa...
Black arrows indicate substantial changes in comparison to Fig 9 (prediction score threshold of 0.9)...