Ok...not a great start. Then comes the contents. So to orient you, [20] is DeOldify.pic.twitter.com/6xWRxdCBXN
Show this thread
So guys...the DeOldify plagiarism story just got hilariously worse. Let me show you what was suggested to me as revisions to the paper to make amends. So, it starts out right away listing me as an "equal contributor", third on the list of authors (LOL):pic.twitter.com/dxoHolnbnh
-
-
-
They start out with the weasel words "inspired by" again. Even though it's basically a carbon copy of DeOldify.pic.twitter.com/h5pIRee4VU
Show this thread -
But then this is where any sort of "gray area" on the truth ends. They then go on to claim that all the major parts that are actually in DeOldify were in fact improvements they made! First: On the use of Spectral Normalization:pic.twitter.com/JXELiTFPGS
Show this thread -
Wrong! DeOldify uses Spectral Normalization exactly for this purpose. Next, it claims that they introduced the use of the "Two Time-Scale Update Rule" as an improvement upon DeOldify.pic.twitter.com/1cHl2qeXwR
Show this thread -
Wrong! DeOldify does that to and the reason why their code does it is because it's freakin DeOldify! Finally, on progressive growing of image sizes- I can't even.pic.twitter.com/GRzN7hcucT
Show this thread -
Here's the DeOldify readme that contradicts all their claims for reference: https://github.com/jantic/DeOldify/blob/master/README.md …pic.twitter.com/bPfhihloIW
Show this thread -
So in summary....apparently the authors thought I'd be a complete idiot and give them the green-light on this because I'd see my name on it and be like "wooo boy, I'm a real academic now!" Nope. Nice try.
Show this thread -
I messed up slightly- DeOldify -used to do- progressive image resize gan training (still does for non-gan). But it certainly doesn't do progressive growing of the architecture and never has!pic.twitter.com/YFSjoWzZD8
Show this thread
End of conversation
New conversation -
-
-
This is just disgusting. Don't give in, you and your colleagues deserve the credit you are due to.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I haven’t read the paper but based on what you have here, the authors don’t have any contributions. Probably, that is why they didn’t mention your work.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.