This program will determine the genre of a given song/songs out of the 5 following:
Metal, Rock, Hip-Hop, EDM, and Pop.
When predicting individual snippets, it had an accuracy of about 55%, however when predicting full songs, this increased quite a lot to about 75%.
I used the Google AudioSet during training - https://research.google.com/audioset/dataset/index.html
And you can find the model I developed here - https://mega.nz/file/oA1WDKjL#_hBK60YVySvi-DKwqnFkS12TrRm8qBD7vpFj3YgNSwA
There's a basic GUI using PyQT5 which will change the ID3 tag of the songs you predict. Alternatively theres the CLI, but I didn't bother implementing it changing ID3 tags or anything like that.
There's also a password menu in the GUI, but the password is set to nothing.
This is my first large project using CNNs and Keras, so the code is a bit all over the place.