linto-ai / linto-desktoptools-hmg

GUI Tool to create, manage and test Keyword Spotting models using TF 2.0
GNU Affero General Public License v3.0
12 stars 2 forks source link

Feature extraction slow #10

Closed StuartIanNaylor closed 3 years ago

StuartIanNaylor commented 3 years ago

Feature extraction on a I5-3570 takes over 10 minutes and longer than training?

Which is no problem but in comparison to what I have been using with the current tensorflow examples seems very slow. https://www.tensorflow.org/tutorials/audio/simple_audio

The above is just spectrogram but if you add the mfcc code to the above its massively faster, is it because mfcc calc is on the fly so overall training is slower via the tensorflow examples?

https://www.tensorflow.org/api_docs/python/tf/signal/mfccs_from_log_mel_spectrograms

Ignore me its the python cache as 2nd run is no comparison.