Replies: 1 comment
-
|
More questions: Is any post-processing applied during the evaluation? If so, how is it done? Is the evaluation one one a frame-by-frame basis, or on a per-event basis (each continious activation)? |
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
-
Hi! Thanks for releasing a performant VAD and also providing evaluation datasets.
Can you provide a bit more details on how the evaluation was done? In order to make it possible for others to reproduce the results.
Some concrete questions:
How did you map the labels of the AVA Active Speaker dataset to binary speech/notspeech? As there are 3 of them: {SPEAKING_AND_AUDIBLE, SPEAKING_BUT_NOT_AUDIBLE, NOT_SPEAKING}.
Is the provided LibriParty dataset a result of following the "official" procedure to generate?
https://github.com/speechbrain/speechbrain/tree/develop/recipes/LibriParty/generate_dataset
Beta Was this translation helpful? Give feedback.
All reactions