The model is simply repurposed from the colorization model. No changes to architecture, and no change in size. A more challenging one:pic.twitter.com/8RjoV95Dh7
Show this thread
Remember that "truly all in one" DeOldify model I proposed? Well here's a few examples from a working model I'm currently training that does it all: Noise removal, contrast/brightness adjust, blur reduction, and of course colorizing. It also restores color photos (not shown).pic.twitter.com/SW6V211MLB
-
-
-
-
This one got livened up a bit. Though it still has weirdness that needs ironed out in training.pic.twitter.com/Jtz281MevW
Show this thread
End of conversation
New conversation -
-
-
I guess Deoldify Algorithm not only uses Neural network but also some image processing is involved? Or the Whole algorithm is driven by Neural Nets?
-
It's 100% a single neural network in this case with the exception of turning black and white images into pure grayscale with Python's Pillow library (to standardize the input).
- Show replies
New conversation -
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
That’s impressive!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
It is incredible how much you’ve progressed this specific field of ML that I think anyone who can’t be excited by this, I believe cannot experience joy
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I wonder about the expressions and poses in old photos. ...Something often unnatural about them. Is it just due to slow shutter speed? Or do you have another theory?
-
I've seen this article before and I think it covers the possibilities pretty well:https://time.com/4568032/smile-serious-old-photos/ …
End of conversation
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.