This stuff was never meant to run more than once and it kind of looks like it.  bktrain.py trains the neural network classifier for detecting barks and pickles it. bkprocessor.py processes audio files using the classifier and visualizes the results.  I've provided a pickled neural network so you don't need to run the trainer, but you'll still need some dependencies (PyBrain, PyGame, PyLab) and a pair of two-channel audio files for input.  When the realtime visualization stops, you'll have a huge pile of PNGs which you can stitch together into a video with something like ffmpeg.

Everything is © Colin McSwiggen.