This attempts to perform nuclei segmentation of histology images using a Conditional Adversarial Network. It uses tf.keras and eager execution to implement the patch-GAN following the example from the tensorflow team: pix2pix_eager.ipynb
This creates an RGB image where the red channel is 1.0 for all "background" pixels; the blue channel is 1.0 for all "nuclei" pixels; and the green channel is 1.0 for all "boundary" pixels. The boundary pixels attempt to describe vertices of a polyline around each nuclei.
See also: Image-to-Image Translation with Conditional Adversarial Networks
U-Net: Convolutional Networks for Biomedical Image Segmentation
