Thursday, July 28, 2016

awesome-audio-cnn 'brain' file

In the last post I mentioned that I was making public a github project that I've been working on. The project is called 'awesome-audio-cnn' and it uses a neural network to select music for automatic playback on a simple server. The project '' file suggests that you train your own neural network file for the server. On reflection I thought this was too difficult.

I have included in the git repository (mentioned in my last post) a file for operating the neural network that is at the heart of the awesome-audio-cnn project.

The file is 23.9M in size, so it will take a long time to download. What this means is that you don't need to train the neural network yourself. It allows you to get a faster start up. I should note that the file was developed at my location, on my audio collection. It may not act as you expect on your files. There is no provision, incidentally, for comparisons on the basis of tempo. The songs you get out of this neural net will not be the same with respect to tempo.

The inclusion of this file may mean that some with slow internet connection or limited disk space cannot use this project.

That said, someone with time on their hands might enjoy setting up this kind of server. The info in the file is provided on an 'as is' basis, without warranties or conditions of any kind.

The file is called 'fp-test.bin' and it is located in a folder called 'acnn-brain', for lack of a better title.

No comments:

Post a Comment