Original image for reference to really drive the point:pic.twitter.com/gjjJ15GVWW
Show this thread
That image on the top right is a great example of how the DeOldify is clearly picking up on some sort of very subtle physical cues on color as opposed to just memorizing objects. Both GAN and non-GAN based approaches converge nearly the exact same solutions. It's not random.https://twitter.com/citnaj/status/1272977889571426304 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
To be clear, if you retrain the models or even change approaches (GAN vs not GAN), that art work will be colored almost identically!
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Love this. Thanks for sharing. More, please!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The small triangle on the right border is clearly a danger road sign. The black square at the center and on the top is clearly a monitor :)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Maybe not random, but still not correct either:pic.twitter.com/ynXpEBGRJA
-
Yeah not claiming correctness. Getting stuff like this actually 100% correct is a super a long tail goal that I don't see being resolved for many years. What makes me happy is the artistic slant of the model creating the illusions.
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.