Brandon Lee

Will you be using French version of Common voice? if so, I think it can possibly work. You will need to find the right french dictionary and replace this line...

it's not merged yet but I recommend using `vocab_stitching` branch

Unfortunately, that feature has been outdated. this version may support it https://github.com/castorini/howl/tree/dba1515acc5c49dc15b594ea97228cb2aedb80a2 it shouldn't be hard to add back the support but I am not personally interested in this direction....

With #66 it is possible to generate enough samples for train/dev/test. The trained model generally reports high accuracy on the generated dev/test However, when I test the detection, I found...

tested both res8 and seq-lstm. I suspected that seq-lstm would be better but both models had low acc that I wasn't able to tell if one is actually better than...

Current experiment plan: train a model only with only with neg + stitched data set. evaluate on real samples

hey_fire_fox should be able to give us the initial results but we might not have enough "real" samples for other wakewords. Unlike common voice, GSC is collected per-word. Therefore, we...

> Stitched datasets ``` train ds - num_examples=27628, audio_length_seconds=116447.25187500067, vocab_counts=Counter({'fire': 5639, 'fox': 5211, 'hey': 5028})) Dev pos dataset - AudioDatasetStatistics(num_examples=2500, audio_length_seconds=3019.3278125000006, vocab_counts=Counter({'hey': 2500, 'fire': 2500, 'fox': 2500})) Dev neg dataset...

To improve the audio sample quality, I have applied secondary filtering with keyword spotting for hey_fire_fox, it has dropped 574 invalid samples while it generates 10000 samples

Keyword Spotting verification definitely helped. However not enough < res8 > > stitched datasets ``` Dev positive:: ConfusionMatrix(tp=878, fp=0, tn=0, fn=1622) Dev negative: ConfusionMatrix(tp=0, fp=0, tn=808, fn=0) Dev noisy positive:...