Inpainting
Generating the contents of an arbitrary region of an image, conditioned on everything around it.
Official source: Context Encoders: Feature Learning by Inpainting →
Harder than local retouching because it demands both halves at once: understanding the whole image, and committing to one plausible hypothesis for the missing part. Trained on a reconstruction loss alone the result blurs; the sharpness associated with the technique comes from an adversarial term that can handle several answers.
Sources 1 source on record · Automated review 3 angles
Sources
What this page is based on — every source is verified, and links out whenever the document is still reachable.