sensein / senselab

senselab is a Python package that simplifies building pipelines for biometric (e.g. speech, voice, video, etc) analysis.
http://sensein.group/senselab/
Apache License 2.0
10 stars 3 forks source link

Task: refactor extract_mel_filter_bank_from_audios #178

Closed satra closed 2 weeks ago

satra commented 3 weeks ago

Description

the current extract_mel_filter_bank_from_audios extraction (https://github.com/sensein/senselab/blob/e5a935a808782cf7ca7d7836b5e0fab45858cd1f/src/senselab/audio/tasks/features_extraction/torchaudio.py#L121) extracts spectrograms for all audios. it would be nice to have that as an optional input in case they have been already generated or broken into separate functions.

Tasks

Freeform Notes

more generally consider optimizing internal parallelization so that scale of number of files doesn't can be handled with limited memory. (this may apply to multiple functions).