How can i get the video and audio pairs of audioset?

[SteveTanggithub]

[YuanGongND]

hi there,

I think youtube-dl is most commonly used to download AudioSet.

Due to copyright reasons, we cannot provide raw audio and videos of AudioSet.

-Yuan

[sukun1045]

Hi Yuangong,

I am wondering whether you could provide the script for downloading the AudioSet via youtube-dl? I tried myself on eval set so far but found some videos were missing/failed to download. It would be great if you could provide the download script if possible so that I could do more consistent comparison in the end. Thank you!

[YuanGongND]

hi @sukun1045,

Thanks for reaching out.

Unfortunately, I do not have that script (I am not the one who downloaded the dataset) and the script might no longer work. FYI, it is not trivial to download AudioSet from YouTube and you need some engineering effort.

I tried myself on eval set so far but found some videos were missing/failed to download

This is actually very common because YouTube videos can be removed with time. But if you can download the majority of the video, that would be fine.

It would be great if you could provide the download script if possible so that I could do more consistent comparison in the end.

We already tried our best to release the ids we used for training and evaluation. If you are only interested in Audio, you should be able to find a direct download link online. I am not aware of a video download link though.

-Yuan

[sukun1045]

Hi @YuanGongND ,

Thanks for your reply. In terms of audio-only, I have yet to find a link that could download all the audio files. It would be super helpful if you could point me to the direct download link.

[YuanGongND]

@sukun1045

Please check this https://github.com/YuanGongND/ast/issues/108#issuecomment-1663723686.

Also please kindly read that entire thread on how to reproduce the results based on that version.

-Yuan

[sukun1045]

@YuanGongND

Got it! Thanks!