Wrong values when using from python #16
Description
Hi, it seems there are some issues with either your code or how I invoke it.
For a file playing piano notes, then converting them to "midi pitch", I get following results for the first 10 (time in ms, pitch):
[(341, 54), (342, 54), (343, 54), (344, 54), (345, 53), (10988, 35), (10989, 35), (10990, 35), (10991, 35), (10992, 35)]
doing the same in the provided matlab script gives me very different values, specifically, pitch 35 is getting recognized from 3752ms on. (The first five notes in the ground truth file are (again: pitch in ms, midi-pitch): 3680,35 3987,35 4304,36 4604,35 4914,41)
In other words, the matlab script correctly finds the first pitch 80ms after the "true" onset (and most others as well), but your script completely fails to recognize most things.
I made sure the params are the same, i.e. pmin=60, pmax=1200, st=0.3, dt=0.001, mel=False
I noticed it seems to be a 24bit wav file, which some libraries can't read correctly - may that be the issue? Although I must admit that would IMO result in an incorrect detection length in the time dimension, which seems to be correct in your code.
If you are interested in reproducing & debugging, I can provide the testfile privately, just leave me an email addy or sth.