This is a Tensorflow 2.0 implementation of DeepDream.
This notebook demonstrates how you can start dreaming up images of your own with no more than a few lines of code.
Enjoy!
- I intended to dream using a network trained on facial images, but ran into issue making forward passes through the net with images. I hope to fix this because I think it would yield interesting dreams.
- I found this cool project done by sprawld that allows the user to stylistically guide the dreaming process by providing an additional image to the model. It seems to yield super cool dreams, so I hope to get that working here as well.
- A Google AI blogpost, Inceptionism: Going Deeper into Neural Networks, helped me understand the DeepDream algorithm at a high level.
- The official Keras implementation of DeepDream by Francois Chollet guided my programmatic understanding of the DeepDream algorithm.
- Josh Gordon's TF 2.0 implementation of DeepDream, though simplistic and lacking in functionality, helped me get familiar with the changes that came with TF 2.0.
- I referenced krasserm's face-recognition project in accessing a trained facial recognition model in TF 2.0.