Image Translation by Domain-Adversarial Training

<div>Our approach learns to “translate” an input image into the one of desired character with or without aligned data: (a-1) image deraining; (a-2) image colorization; (a-3) image inpainting; (b-1) sketch to photo; (b-2) image style transfer; and (b-3) gender swapping.</div>

Computational Intelligence and Neuroscience

fig1

Figure 1

Figure 1: Image Translation by Domain-Adversarial Training 