DeepAudioClassification

Why we included this project

This project is a small, honest example of how a music genre classifier gets built from raw files. You drop labeled MP3s into a folder, slice them into short chunks with sox, train a TensorFlow/TFLearn model, and check the test accuracy. The model lives in model.py and most knobs sit in config.py, so tweaking and rerunning is straightforward. It is not a finished product: the author admits the step that labels new songs is not wired up, but points to the functions and eyed3 calls needed to finish it. For anyone learning audio classification, that unfinished edge is what makes it useful, and the Medium write-up walks through the same approach.