DeepDream Hack Night
Details
We will get together and walk through how to create our own "Deep Dream" images like the ones you might have seen floating around the internet earlier this month. They are terrifying and awesome and have an uncanny resemblance to psychedelic (http://www.citylab.com/design/2015/07/psychedelic-maps-made-with-googles-deepdream-code/397961/) visions! Google open sourced their framework for creating these images (http://googleresearch.blogspot.co.uk/2015/07/deepdream-code-example-for-visualizing.html), so we will work through their scripts, and then feed in our own images.
https://lh3.googleusercontent.com/0bfSmR-nblU3ow4-WMsN7rvnNDOKMoIIi2LH2_lMRRA=w716-h448-no
Check out the gallery (https://photos.google.com/share/AF1QipPX0SCl7OzWilt9LnuQliattX4OUCj_8EP65_cTVnBmS1jnYgsGQAieQUc1VQWdgQ?key=aVBxWjhwSzg2RjJWLWRuVFBBZEN1d205bUdEMnhB) for more examples and/or the #DeepDream hashtag on Twitter (https://twitter.com/search?q=%23deepdream&src=typd).
More info:
The code is based on Caffe (http://caffe.berkeleyvision.org/) and uses available open source packages, and is designed to have as few dependencies as possible. It would be helpful to install at least the Python packages prior to the meetup, but we will walk through all the steps to make sure no one is left behind. To get started, you will need the following (full details in the notebook (https://github.com/google/deepdream/blob/master/dream.ipynb)).
• NumPy (http://www.numpy.org/), SciPy (http://www.scipy.org/), IPython (http://ipython.org/), or a scientific python distribution such as Anaconda (http://continuum.io/downloads) or Canopy (https://store.enthought.com/).
• Caffe (http://caffe.berkeleyvision.org/) deep learning framework (Installation instructions (http://caffe.berkeleyvision.org/installation.html))
• Also, check out my own installation instructions (https://github.com/wimlds/deepdream-workshop) for getting this all to work on OS X (with Homebrew Python).
Once you’re set up, you can supply an image and choose which layers in the network to enhance, how many iterations to apply and how far to zoom in. Alternatively, different pre-trained networks can be plugged in.
Note: As of right now, there will not be food, but feel free to bring take-out if you want to eat at the hack night.