2f metric

[faroit]

Hi @Ashvala! this is a great project, finally someone tackles this! :-)

Do you plan to support the 2f metric directly here that you mention in the paper?

[Ashvala]

Hey! Thanks a lot! I will work on it during the christmas break :)

[faroit]

@Ashvala awesome, ping me here if you need help or a code review

[faroit]

@Ashvala is there any news on this? I think the community would love to have a python 2f model that works out of the box :)

[Ashvala]

@faroit Hey! Sorry about the lack of progress on this - I hope to get back to working on this sometime this week.

[halil-erdogan]

Hey @Ashvala,

Thanks for great project! Are there any progress on this?

[Ashvala]

@faroit It's almost been a year! I'm so sorry. It's been a whirlwind of a year! I've finally gotten around to implementing 2f in some shape here: #9

Once the package installs, you should be able to import PEAQ2F through aquatk.metrics.PEAQ.PEAQ2F.

You should be able to do:

peaq_2f = PEAQ2F()
peaq_2f.analyze_files(a, b)

or

peaq_2f.analyze_arrays(a, b, sample_rate=<sample_rate>

My worry with this implementation is that I used the coefficients described in https://www.audiolabs-erlangen.de/resources/2019-WASPAA-SEBASS. Since I used a different implementation of PEAQ as my basis , I'm seeing values of over a 100, which if I understand the metric correctly, is not appropriate. Therefore, i have manually clamped the score for now.

Please let me know how I can improve the implementation to better suit your needs :)

@hedonistrh Looks like I finally have time to work on this, so, progress is slowly, but surely going to happen :)

Once again, I'd like to apologize for the protracted delays with this.