I’ve been playing with Google Deep Dream again lately and ran across a very nice Docker container that makes it simple to test out with no extensive setup. The only requirement is Docker. I decided to add some new functionality to allow you to specify the number of iterations, the scale, and which model to use. My fork can be found here: https://github.com/jbirchfield-ps/deepdream-docker. I have submitted a pull request, but it is not merged yet. Here is how you run it:

docker run -i -t -e INPUT=your_file.png -e ITER=20 -e SCALE=0.10 -e MODEL='inception_3b/5x5_reduce' -v /path/to/your/folder:/data birchsport/deepdream-docker

//example (ITER, SCALE, and MODEL are optional)
docker run -i -t -e INPUT=matgrix.jpg -v /Users/jbirchfield/Pictures:/data birchsport/deepdream-docker

Here is the original picture before the dream sequence: devil

Here is an example image after only one iteration: Karellen

And here is the picture after a full 100 iterations: 0099