@AlecRad looks very convincing! Loving those indecipherable artist names and album titles :) What does the model look like? conv. VAE? DRAW?
-
-
-
@sedielem conv GAN - all strided so no explicit pooling steps. Trying one more project then planning a write-up on these types of models! - Voir les réponses
Nouvelle conversation -
-
-
@AlecRad this is really cool. Can the model generate arbitrary larger size samples after it is trained? -
@hardmaru Thanks! Not currently possible, just added another upsampler to the previous architecture to increase the resolution. - Voir les réponses
Nouvelle conversation -
-
-
@AlecRad@whitequark 4th from left, 3rd from top: BongMerci. Twitter en tiendra compte pour améliorer votre fil. SupprimerSupprimer
-
-
-
-
@iangilman unfortunately the results of generative music models aren't nearly as good yet ;)
Fin de la conversation
Nouvelle conversation -
-
-
@AlecRad I would try to enforce better matching color schemes, e.g. http://www.tigercolor.com/color-lab/color-theory/color-theory-intro.htm … (not sure if it's possible to do here though) -
@rid5x It's probably possible since the framework's general - another thing would be working in lab and converting back to rgb maybe?
Fin de la conversation
Nouvelle conversation -
Le chargement semble prendre du temps.
Twitter est peut-être en surcapacité ou rencontre momentanément un incident. Réessayez ou rendez-vous sur la page Twitter Status pour plus d'informations.