search

Ant-Brain / EfficientWord-Net

OneShot Learning-based hotword detection.
https://ant-brain.github.io/EfficientWord-Net/
Apache License 2.0
215 stars 34 forks source link

Reduce False Positives #8

Closed Siyamfahad closed 2 years ago

Siyamfahad commented 2 years ago

I really love this project... working well except one thing its having too much false positive, how can I reduce it?

aman-17 commented 2 years ago

Change the threshold value.

Siyamfahad commented 2 years ago

sorry, i didn't understand can you please explain it little?

aman-17 commented 2 years ago

Go to the engine.py file and in line 21, threshold:float=0.9 increase the threshold value. If the model is giving false positives again, provide a few more samples of hotwords.

Siyamfahad commented 2 years ago

@aman-17 thank you so much :) can I increase the threshold by more than 1 sir?

aman-17 commented 2 years ago

No, you cannot. Please read the documentation or preprint before using our model. 👍🏻

TheSeriousProgrammer commented 2 years ago

@aman-17 thank you so much :) can I increase the threshold by more than 1 sir?

the confidence ranges only between the range of 0 to 1 => 0 to 100% One method to reduce false positives could be is to feed recordings from you own voice

if the results are satisfactory feel free to share a video of your implementation and close the issue

TheSeriousProgrammer commented 2 years ago

closing due to inactivity ,feel free to open back