The reason deep dream does not generate visually interesting results is because it uses a network trained on ImageNet (dogs and birds...)
I've updated the Keras deep dream example to be closer to the original, but the hyperparameters still need tweaking https://github.com/fchollet/keras/blob/master/examples/deep_dream.py …pic.twitter.com/MhfewcVG7N
-
-
-
Training e.g. an artist classification network on thousands of paintings would allow to generate much more artistic and creative dreams
- Show replies
New conversation -
-
-
@fchollet on a related note, it would be great if keras.applications provide pretrained models on mit places and pascal datasets. :)Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What is that?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
what hyp params u used for this image?
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.