Synthetic images aid the recognition of human-made art forgeries | Hacker News Reader

Synthetic images aid the recognition of human-made art forgeries

arxiv.org

29 pointsevanb8 comments

If the synthetic images are not classified as forgeries, does that mean they're essentially generating model-specific fuzzed versions of the originals? So rather than just adding noise to the pixels, it's able to pentrate the noise deeper?

That sounds generally useful with deep models.

I posted this more than 6 hours ago, no matter what the timestamp says. This has happened to me before [0] and I still wish the time weren't rewritten.

https://news.ycombinator.com/item?id=29529099

It's so important to check when one is parking one's time machine.

If it’s helpful, in my mind I am personally putting one extra credit tally next to the screen name evans.

It's not about credit; this thread shows that I submitted it. It's about saying I did something at a time when I was asleep.

Sorry. Unfortunately that section of my brain is immutable so I am unable to strike it out.

That's an artifact of HN's second chance pool system. The system is described at https://news.ycombinator.com/item?id=26998308 and links back from there. About the timestamps, there are past explanations here: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que....

Sorry for the confusion—I know it's weird but if we don't do it, then the threads fill up with other comments about timestamps, like "why is this old post ranking alongside all the new ones". The current system is imperfect but seems to be the available optimum.

I see. A tricky design constraint; if you _indicate_ it's from the second-chance pool it might not get a fair shake but if you don't it'll look old and unpopular and might not get a fair shake. Maybe for something like that an on-hover/tooltip on the timestamp might make sense, but I appreciate the difficulty and your chiming-in to let me know what's going on!

Reply on news.ycombinator.com