Figure 2.

Cropped images of typical brightfield images, ground truth fluorescent, and predicted channels from the test dataset (Table 1) for the U-Net and cWGAN-GP models. Presented at scale of the input and output size used in the model networks (256 × 256 pixels), the first column contains the three brightfield z-planes which form the input, and the subsequent columns are the five fluorescent output channels. The stitching method has been applied to the model predicted images, and small artefacts of this process are visible. Images are independently contrasted for visualization.