Table 1.

Intersection over Union metrics of U-Net-based networks trained on the light optical microscopy data set for different model initializations and downscaling factors.

#	Model	Model initialization	Downscaling factor	IoUbg	IoUfg
1	Vanilla U-Net	Random	native	87.0 ± 1.5	70.2 ± 1.2
2	Vanilla U-Net	Random	0.5 × 0.5	87.2 ± 1.9	69.5 ± 1.8
2v	Vanilla U-Net	Random	0.5 × 0. 5*	86.8 ± 1.5	69.8 ± 1.3
3	U-Net VGG16	Pre-trained	native	87.6 ± 0.9	71.3 ± 1.7
3v	U-Net VGG16	Random	native	86.3 ± 1.3	69.3 ± 1.5
4	U-Net VGG16	Pre-trained	0.5 × 0.5	87.1 ± 1.7	71.6 ± 1.7

The superscript ^v indicates a validation experiment conducted to test a particular hypothesis, as described in the section “Segmentation results”. The superscript * indicates that downscaling was performed after tiling.