1. The encoder & decoder 2. The VAE (with a custom train_step) 3. Training itpic.twitter.com/J8D4MjZWpt
Show this thread
How to implement & train a convolutional variational autoencoder with Keras in under 50 lines of code: https://colab.research.google.com/drive/1COeSoasrpY-w3uF1JPVksLEjx4Tx2ns5 …pic.twitter.com/MK580AGH53
-
-
-
Here's a slightly tweaked version that does a better job at showing the latent manifold of MNIST digits (I'll put this one in the next edition of my book to illustrate the manifold hypothesis)pic.twitter.com/3bx5LYwaNf
Show this thread
End of conversation
New conversation -
-
-
Thank you so much for sharing .
@fchollet.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I’m a user of Keras since 5 years and seeing the serie of Colab notebooks is so great. Thanks! Do you have a list of those notebooks?
-
They will all be on http://keras.io soon (as soon as next week, maybe)
End of conversation
New conversation -
-
-
Building a VAE for 256 x 256 grayscale images. Are there a sizing rules to thumb for size of latent space, parameters (network size, breadth, depth) and a good weight initialization strategy? This is tricky and networks quickly converge to very suboptimal reproductions.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thank you so much
@fchollet . Your Deep learning book is by far the best I have read! I have a question on autoencoders: could you include any example with non-mnist images. ThanksThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Nice. Had to upgrade tensorflow from 2.1.0 to 2.2.0-rc4 to make it 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.