Here’s the video that was used for training, a (not so perfectly) looping style transfer. The actual video is 1200x1200 px so the pix2pix model is randomly cropping 1024x1024 px bits out of it.pic.twitter.com/fu32dE98pc
-
-
Prikaži ovu nit
-
The first model learns to capture the movement of the video while the second model tries to keep some kind of structure. Here is an example of running just the frame prediction model. The effect is a lot more scrambled movement.pic.twitter.com/TpNr7zRf9z
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.