Issues for the HAMR Sprint, 2016-08-06

[Tagging @craffel @ejhumphrey @justinsalamon @stefan-balke – feel free to mention anyone else that might be attending HAMR and has some interest in any of the topics below]

I’d like to do a sprint on librosa for HAMR this year. The idea here is to get people together to work on issues slated for the next release (0.5), or help new developers ease into contributing.

Here’s a list of some issues that I think we could reasonably work on over the course of a day, rated by estimated difficulty and the amount of existing preliminary work done so far.

• 😄 [ ] #360 : trimming and splitting
• 😄 [++ ] #305 : time-reassignment
• 😄 [ ] #278 : logo
• 😄 [+ ] #359 : API change to hpss with masks
• 😄 [++ ] #157 : adding numba support to tight-loop functions
• 😓 [ ] #18 : refactoring filterbank construction
• 😓 [+ ] #318 : API change to feature normalization
• 😓 [++ ] #297 : salience function
• 😰 [+++ ] #165 inverse CQT
• 😰 [ ] #70 #128 #140 various related IO issues
• 😰 [++ ] #44 #269: variable-tempo beat tracking and API changes
• 😰 [ ] #303 : local contrast normalization
• 😰 [ ] #12 / #394 : improved chroma and log-frequency representations
• 😰 [ ] #391 : Refactor tests

If/as people volunteer for certain tasks, I’ll attach names to the issues above.

[lostanlen]

I’d like to claim #18.

[stefan-balke]

@stefan-balke what did you have in mind for “reassigned spectrogram”? Something different from the existing ifgram?

No, it would take the ifgram and “resort” the frequencies depending on a target frequency axis (might be from lin- to log-frequency). It should be straightforward and I guess painfully slow but for the resorting we might as well add numba.

Basically it’s all described here: http://perso.ens-lyon.fr/patrick.flandrin/0065-Ch05.pdf