bmcfee / laplacian_segmentation Goto Github PK

I am working on the music segmentation problem and I have a few questions about your code when I studied it. It would be great if you could give some pointers.

The function label_entropy in segmenter.py returns the entropy of the "labels" but not the hits calculated inside the function. Is this intended? Then, if it was the "hits" that was intended to be used.
Wouldn`t it always return 1.0 after divided by log(k)? Because of the np.unique function only returns the unique value of an array, it is somewhat effectively the same as range(k) in some sense. Then the next for loop in the function will just fill hits with one count for each bin.

Please let me know if I misunderstood anything or missed some steps you
took in the code. Thank you so much!!!

Do i get the structural parts of music like intro, chorus, verse, bridge and outro.
Thanks!

Hi!
I was just trying out some of your code and I noticed listening to y_perc and y_harm that they sounded switched. And indeed, in your current documentation of librosa you have them in the reverse order that they appear on line 131 of segmenter.py. Perhaps they got switched in a newer version of librosa?

Anyway, maybe this doesn't matter anymore, but I figured I'd flag it just in case.

Thanks for the great software!
Sincerely,
Chris Tralie