search

orasanen / ALICE

Automatic LInguistic Unit Count Estimator (ALICE)
44 stars 11 forks source link

Is run_ALICE.sh running SAD ? #14

Closed macramole closed 3 years ago

macramole commented 3 years ago

I have a question regarding ALICE pipeline.

Line 43 and 44 of run_ALICE.sh reads:

# Run SAD on the files
python3 prepare_data.py $THISDIR $DATADIR/

But after reading _preparedata.py it doesn't look as if it is running any SAD related task.

am I missing something ? does voice-type-classifier run SAD ? it is non mandatory to run SAD before voice-type-classifier ?

Reading the code in voice-type-classifier/apply.sh it runs pyannote-audio with mlt parameter which doesn't seem to be running SAD either.

Best regards, Leandro Garber

orasanen commented 3 years ago

Oh, right. The comment can be somewhat confusing. The "SAD" is done jointly on row 49 with speaker diarization, so the voice-type-classifier is responsible for both speech detection and speaker attribution.

bash $THISDIR/voice-type-classifier/apply.sh $THISDIR/tmp_data/ "MAL FEM" --device=$GPU 2>&1 | sed '/^Took/d'

I have now fixed the comment to be clearer.