[ASR] Added a script for evaluating metrics for audio-to-audio

[anteju]

Added a script for evaluating audio-to-audio metrics for a manifest file (audio_to_audio_eval.py)

Signed-off-by: Ante Jukić [email protected]

What does this PR do ?

This PR adds a script for evaluation of audio-to-audio metrics.
It is similar to speech_to_text_eval.py

This scripts depends on the process_audio.py script and inherits all its arguments.

Collection: ASR

Changelog

• Added audio_to_audio_eval.py
• Minor change in proces_audio
• Added an option to get_full_path to specify data_dir directly instead of using dirname(manifest_file)

Usage

• You can potentially add a usage example below

To score a dataset with a manifest file that contains the input audio which needs to be processed and target audio

python audio_to_audio_eval.py \
    model_path=null \
    pretrained_model=null \
    dataset_manifest=<Mandatory: path to a dataset manifest file> \
    output_dir=<Optional: Directory where processed audio will be saved> \
    processed_channel_selector=<Optional: list of channels to select from the processed audio file> \
    target_key=<Optional: key for the target audio in the dataset manifest. Default: target_audio_filepath> \
    target_channel_selector=<Optional: list of channels to select from the target audio file> \
    metrics=<Optional: list of metrics to evaluate. Defaults to [sdr,estoi]>
    batch_size=32 \
    amp=True

To score a manifest file which has been previously processed and contains both processed audio and target audio

python audio_to_audio_eval.py \
    dataset_manifest=<Mandatory: path to a dataset manifest file> \
    processed_key=<Optional: key for the target audio in the dataset manifest. Default: processed_audio_filepath>
    processed_channel_selector=<Optional: list of channels to select from the processed audio file> \
    target_key=<Optional: key for the target audio in the dataset manifest. Default: target_audio_filepath> \
    target_channel_selector=<Optional: list of channels to select from the target audio file> \
    metrics=<Optional: list of metrics to evaluate. Defaults to [sdr,estoi]>
    batch_size=32 \
    amp=True

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

• Related to # (issue)

[jbalam-nv]

LGTM