search

YuanGongND / ssast

Code for the AAAI 2022 paper "SSAST: Self-Supervised Audio Spectrogram Transformer".
BSD 3-Clause "New" or "Revised" License
365 stars 61 forks source link

ssast pretrained solely on Audioset-2M #1

Closed Moadab-AI closed 2 years ago

Moadab-AI commented 2 years ago

In the paper there is a section where you compare the model trained self-supervised only on Audioset-2M to the default model where you combine it with librispeech. Could you please kindly share the self-supervised pretrained model on Audioset-2M alone?

Thanks in advance.

YuanGongND commented 2 years ago

Thanks.

I will need some time to check where the model is on our server and convert it to the release version (the model definition is simplified in the released version). I will keep you posted.

Best, Yuan

YuanGongND commented 2 years ago

Ok, I add all models in Table 2 of the paper. Please check the readme file again.

-Yuan