Skip to main content
. 2021 Oct 27;1(7):100105. doi: 10.1016/j.crmeth.2021.100105

Figure 4.

Figure 4

Visualization of segmentation results from U-Net and VGG19D-U-Net trained on the phase contrast microscopy dataset

(A and B) Edges extracted from the ground truth mask and predictions from U-Net and VGG19D-U-Net are overlaid on the first frame of the movie of a PtK1 cell acquired by a phase contrast microscope. Each edge is represented by one of three primary colors. The overlap of two or more edges is represented by the combination of those colors.

(C) Progression of cell edges segmented by U-Net and VGG19D-U-Net overlaid on the first frame of the movie of a PtK1 cell acquired by a phase contrast microscope (blue, 0 s; red, 1,000 s time points). Same representative image as shown in Figure 1B. White dashed boxes indicate regions that either U-Net or VGG19D-U-Net segmented incorrectly, or both models segmented incorrectly. Bars: 32.5 μm.