Putting Image generation through its paces with the latest VQGANs from Heidelberg University.
All images were generated entirely offline and prompts were made up on the spot with the help of a random word generator. Some results are stranger than others.
While this example only showcased VQGAN ImageNet (f=16), 16384 , I adapted VQGAN OpenImages (f=8), 8192, GumbelQuantization as well. Unfortunately the reduction in clarity and performance led to poor results so they were left out of the video.
Special thanks to the folks behind Taming Transformers and Advadnoun for their excellent documentation!
Link to more information about VQGANs and their process:
https://compvis.github.io/taming-tran...
introduction: (00:00)
okay test: (2:27)
thats a tough one: (2:48)
Reese's cereal: (3:07)
Scissors on the beach: (3:22)
a bird chair: (4:00)
a pizza crayon: (4:18)
a metal person: (4:37)
a beautiful vacation: (5:55)
Sur cette page du site, vous pouvez voir la vidéo en ligne Speed Testing Image Generating VQGANs durée heure minute seconde en bonne qualité , qui a été Téléchargé par l'utilisateur Caz Czworkowski 27 juillet 2021, Partagez le lien avec vos amis et connaissances, sur youtube cette vidéo a déjà été regardée 236 fois et il a aimé 13 téléspectateurs. Bon visionnage!