Spoken-language-identification icon indicating copy to clipboard operation
Spoken-language-identification copied to clipboard

Published 20 hours ago • verified publisher icon YerevaNN

how to predict a new audio?

Open 363325971 opened this issue 7 years ago • 1 comments

I finished training and produced a state file. At this point how to predict a new audio file? I tried --mode test, however it request the audio file had been labeled.

363325971 avatar Mar 23 '17 09:03 363325971

There is no special procedure to predict the language for a new audio file. Now in order to do that you must create new listfile containing only that audio file, substitute that listfile with test set listfile and then run the model with --batch_size=1 and --mode=test.

I will be very glad to accept a pull request which makes prediction on a single audio file easy.

hrayrhar avatar Mar 24 '17 18:03 hrayrhar