GANs for image segmentation

I wonder if you could use GANs to generate image segmentation masks? What do you think? I would create a Generator that tries to build images with masks on it (coloring the pixels within the image). The discriminator decides if this image is created or a real image mask.
Afterwards checking a pixel values against the mask color returns the mask coordinates for the image?

Hast anyone tried a similar approach to image segmentation?

1 Like