Audio clips were turned into spectrograms and heavily masked to make a non-trivial training task for an auto-encoder.
Meta’s masked auto-encoder for audio was trained on heavily masked data, and was then able to reconstruct audio files with impressive fidelity.
Learn more: https://spectrum.ieee.org/unsupervise...
On this page of the site you can watch the video online Masked Version of Training Data for Meta AI's Masked Auto-Encoder with a duration of hours minute second in good quality, which was uploaded by the user IEEE Spectrum 30 September 2022, share the link with friends and acquaintances, this video has already been watched 1,631 times on youtube and it was liked by 3 viewers. Enjoy your viewing!