Opus audio decoding support

[vadimkantorov]

Is opus codec (in ogg container - typically with .opus file extension) supported by DALI in audio decoding?

https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/audio_processing/audio_decoder.html says that "procedure can be used for most of the well-known digital audio coding formats as well", and opus is very popular as it's produced by webrtc streams.

https://docs.nvidia.com/deeplearning/dali/user-guide/docs/operations/nvidia.dali.fn.decoders.audio.html says that "It supports the following audio formats: wav, flac and ogg". Does it mean ogg-container/format or ogg-codec? The wording is a bit ambiguous.

Thank you!

---

My original usecase: using https://github.com/triton-inference-server/dali_backend/ and DALI for performant audio decoding in Triton

[jantonguirao]

Yes, DALI should support both OGG Vorbis and OGG Opus for audio decoding. I agree that the wording in the documentation is unclear. Thank you for bringing this to our attention; we will update the documentation to clarify this.

Please let us know if you find any issues with it.

[jantonguirao]

Here is the PR for the documentation changes

[vadimkantorov]

Also maybe worth dropping vague "most well-known audio codecs" wording

E.g. are mp3/aac/m4a/mka supported? Even if not supported - better list all supported codecs/containers (and maybe even provide a battery of example supported audio files) - and if a popular codec is not supported, better also list it explicitly, as people will have questions about codec support matrix anyways...

[vadimkantorov]

For that example speech commands example docs - maybe best put there a hyperlink to the codec support matrix? (As if you evolve it, this place would need to be modified as well - and it's easy to forget such stuff)