in bin/import_cv2.py,
there is a parameter, named MAX_SECS.
if I set to higher value,
then how about the performance of Deep Speech?(for example in case of short audio)
what is a good value for this?
If you read the code, you will see it’s just to filter out too long audio samples, in an effort to avoid OOM and maximize batch size.
yes, I check it.
Does it affect the performance of speech recognition?
what? can you clarify your question?
I confused how RNN based speech recogntion framework can deal with the varialble length of sequence.
I check and study the code, but still I am confusing.
is there a chunking??
or our RNN based framework can deal with the variable length of sequence??
The RNN can handle variable length inputs.