Is vad necessary?

[NemoCpp]

Hi Yuji Tokozume and other authors,
I'd like to try your algorithm on my own dataset for audio classification.
I have quickly read your code, and notice that it seems there is no VAD (voice active detection) processing.
For my own dataset, the audio contains many silence parts.
So, I'd like to ask, is vad necessary for your algorithm or not ?

[Tokozume]

Hi, thank you for your question.
Our algorithm can work without using VAD as well as the standard learning, but we recommend you to use VAD if there are many silent parts in the dataset. We didn't use VAD, because there are not many silent parts in the datasets we used and VAD is not generally used in other papers.