Brian McFee
Brian McFee
In implementing #253 , I ran into this question around the semantics of the `track.split` attribute and what its limitations are/should be. As far as I understand it, splits at...
In setting up mirdata for a new environment, I was unable to run the tests by following the instructions. In this specific case, I had not installed `music21` because it...
Some experiments on the SPAM dataset have turned up a few quirks in the Laplacian segmentation method that shouldn't be too hard to fix. Problematic tracks include: - SALAMI_478 -...
One of the common questions that I get is "how should i build an audio classifier?" My canned response is "openl3 + sklearn RandomForest". The tutorial in the docs is...
Shipping conda(forge) packages would be helpful for having a slightly smoother installation process with dependency on pysoundfile/libsndfile. I'm happy to help get this set up if yall are interested.
**Is your feature request related to a problem? Please describe.** Gluing a matplotlib animation as a `glue:figure` directive doesn't work, in that no animation content is rendered in the final...
#1 lists several deformations left to implement here, and we currently rely on some ad-hoc constellation of backends to implement what we have now. It looks like [pedalboard](https://github.com/spotify/pedalboard) is on...
Our dependencies and tests are a bit out of date at this point, and we should do the following: 1. bump librosa up to 0.7 (at least) 2. upgrade pytest...
Simple(ish) deformers (many from audio degradation toolbox): - [ ] Low-, band-, high-pass filter - [ ] bit-crushing - [ ] companding+quantization noise - should wait until https://github.com/librosa/librosa/pull/1022 is finished...
The current implementation (implicitly) supports only one form of edge behavior: zero-padding. With a bit of work, we could support the usual array of edge modes: - constant padding (non-zero)...