Deep image prior 'learns' on just one image

Re:Deep image prior 'learns' on just one image by amelius on Friday December 01, 2017 @05:31AM (#15823072)

That's something different: a demonstration of an "inpainting" problem. The black pixels were specifically marked as corrupt. Also, the pattern on the cloth is much less granular than noise.

[ Parent | on HN ]

Re:Deep image prior 'learns' on just one image by rsp1984 on Friday December 01, 2017 @11:27AM (#15824864)

Ah, correct. Since the image was in grey I thought is was added noise, not missing pixels. Thanks for pointing that out.

[ Parent | on HN ]

Re:Deep image prior 'learns' on just one image by LoSboccacc on Friday December 01, 2017 @06:48AM (#15823287)

still the problem space is greatly reduced - all we need now is a nn to detect corrupted pixels

[ Parent | on HN ]

Re:Deep image prior 'learns' on just one image by amelius on Friday December 01, 2017 @07:03AM (#15823335)

We need more than just the information where corrupted pixels are.

For example, consider an image of a woman, severly blurred such that her individual fingers cannot be discerned but you can still see that it's a woman. A "learned" context-aware approach can now deduce that the "blob" that is her hand should be inpainted by fingers. A non-learned approach can't do that.

[ Parent | on HN ]

Re:Deep image prior 'learns' on just one image by asfdsfggtfd on Friday December 01, 2017 @07:33AM (#15823432)

But the fingers it draws will be random fingers. Not her fingers.

[ Parent | on HN ]

Re:Deep image prior 'learns' on just one image by LoSboccacc on Friday December 01, 2017 @10:51AM (#15824521)

true but it's still pretty amazing that those specific number fluctuations represent fingers (random or otherwise) without training

[ Parent | on HN ]