deepdream layers using googles places model

Best viewed in 1440p :D, Googlenets cnn places is run for 500 steps on the same image, and the result for the layers: inception3b, output, inception4a, output, inception4b, output, inception4c, output, inception4d, output, inception4e, output is made into an animation and displayed side by side, the places model was used to avoid the abundance of dogs using the standard deepdreaming model.