@citnaj Hi Jason. Thanks so much for DeOldify. It's absolutely amazing. Just to confirm, because it seems like there must be a mistake in the GitHub repo, that DeOldify is released under the MIT license and we can offer a commercialized version?
-
if we can monetize well -- still far from certain, of course -- we'll need to revisit this anti-donation position. you deserve something for this wonderful research. :) any tips on first-timers digging into de-oldify?
I'd say- I have yet to find any sort of image preprocessing that works reliably enough to use in production- so don't bother with that. Higher resolution/higher detailed images == better results, generally. And render_factor 35 is the best default for overall reliability.
-
-
what do you mean by image preprocessing that isn't reliable in production? do you mean a preprocessing service? or using custom models to remove noise, to crop, etc?
-
I mean things like removing noise, auto-contrast, sharpening, etc. I have yet to find a one size fits all solution. What winds up always happening is that it works for a subset of images and you make others worse. A net negative.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.