Here's a shocker: Yes, I'm aware of the excessive purple and other misplaced colors in DeOldify. There's no shortage of people who can't wait to inform me about this. Either make a better model. Or just carry on and trust me that I've thought plenty about it over two years..
-
Show this thread
-
-
Replying to @PulkitRega
Jason Antic Retweeted Jason Antic
Jason Antic added,
Jason Antic @citnajReplying to @Ldavid65You can view it as a bug of sorts. It happens when the model comes across the color of something pretty much unknowable (like most clothing), and so the model basically throws its arms up and says "I give up..purple!". It's a very average color so it works to cheat the tests.1 reply 0 retweets 2 likes -
Replying to @citnaj @PulkitRega
Wouldn't there be a way to make the model more knowledgeable about unknowable things like clothing? Like a new ML project that becomes a clothing detector and inject it in the model? Greetings, an ML nitwit.
2 replies 0 retweets 1 like -
Replying to @Bacon_Sec @PulkitRega
2/ So from that perspective it's a bit of a Quixotic quest to try to train a model for "all the clothing knowledge". But anyway- you pretty much have that as best as you can probably get anyway integrated into the singular color model by training it on diverse examples already.
1 reply 0 retweets 1 like -
Replying to @citnaj @PulkitRega
Do I understand correctly then that the model does know and recognize something is clothing; it just always defaults to the same colour? Or is it really more like it doesn't know what it is, so therefore: purple.
1 reply 0 retweets 0 likes
Probably a little of both going on there since clothing is so varied.
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.