Audio/Video data augmentation

[YUCHEN005]

Hi Authors,

Thank you for good sharing. Since you have added noise on input speech with 0.25 probability, I wonder if you have deployed data augmentation on the input image? (like the image flip that CV community usually do) If yes, how can I use it on the avhubert?

Thank you~

[chevalierNoir]

Hi,

Yes. We do data augmentation on images, including horizontal flipping and random cropping. You can find details here.

[YUCHEN005]

See it, thank you~