Audio thumbnails/summaries in the embedding visualization about tensorboard HOT 5 CLOSED

Hi @act65 and @PFCM,

This is a cool idea. Unfortunately, we don't plan on adding this sort of functionality, but I'd be happy to review and accept a contribution if you decide to contribute.

Regarding implementation, it will be a decent-sized change. You will need to encode the sound url/data in PointMetadata. I recommend using a reserved key for this data (e.g. __AUDIO__).

Then you need to register a custom hoverListener here if metadata of type __AUDIO__ is found in the data. Then on hover you need to render/play the audio using html.

Feel free to re-open this issue when when/if decide on sending us a contribution.

Thanks!

from tensorboard.

Hi @dsmilkov,

I was looking for how to play audio in the TensorBoard's embeddings tab and found this issue.

You already provided a few useful hints for implementing this two months ago. I was wondering if anything has changed since then?

If not, I'd give this a shot, following your pointers.

from tensorboard.

Hi! Nothing changed on our side. Feel free to go ahead with implementation. I'd be happy to review!

from tensorboard.

That's cool.I use embedding to visualize phoneme if has a function like this will be very powerful.
Like this cool project!

https://experiments.withgoogle.com/ai/bird-sounds

from tensorboard.

https://github.com/kokimame/tensoba is a fork with basic audio playback capabilities. No affiliation, just found it looking for a solution :)

from tensorboard.