Replies: 1 comment
-
This metrics look fairly good overall, and in my experience should lead to a reasonably well performing model (with perhaps a higher than desirable false-positive rate). What are the conditions where you are seeing the model miss activations? Certain types of background noise, or certain ways of speaking the wake word? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hey everyone, I am trying to train a model that should detect "Hey Mizu". I have trained 50+ models in order to achieve good results but sometimes my models miss keywords. I think that it is related to my recall score.
My best model's hyperparameters and scores are:
Hyperparameters:
Keyword: Hey Mizu
Number of samples: 50k
Negative weight: 5000
Step Size: 700k
Results:
Accuracy: 0.84
Recall: 0.68
FP rate: 5
Any ideas how can I improve my recall score ?
Beta Was this translation helpful? Give feedback.
All reactions