Conditional Generative-Adversarial Network for inverse Holography

Output of a conditional convolutional GAN trained with spectral normalization on light-field/holography pairs. The generator learns the conditional distribution of holograms (grey, left panels) given a certain light field (red colours on the right).