Questions about model input size and output size

ncarraz / AFILM

MLSP 2021 - Self-Attention for Audio Super-resolution - Keras implementation

MIT License

28 stars 8 forks source link

Questions about model input size and output size #5

Closed bnbn860904 closed 10 months ago

bnbn860904 commented 2 years ago

Dear ncarraz

I'm currently learning about Audio Super Resolution. I want to ask you a question. Why does the output of the model have a length of 8192 instead of double, four or eight times the length of the input ? According to my understanding, to increase the resolution, the length of the output should be longer.

If you can answer me I will thank you very much. thanks.

ncarraz commented 2 years ago

Hello, The number 8192 I am referring in the paper is the input patch length. The output length depends on the input and the scaling factor.

bnbn860904 commented 2 years ago

Got it, thanks for your quick reply.