about GMAD test toolkit / integration

[GuoanZou]

Hi,Thank you for the great work on IQA-pytorch.

I'm interested in using the GMAD (Group MAD) test toolkit for evaluating IQA models. Could you please clarify whether GMAD is supported or integrated in this repository?

If not, do you have any recommendations on how to incorporate GMAD-based testing with IQA-pytorch? Or is there a separate toolkit you would suggest for GMAD evaluation?

Any guidance or pointers would be greatly appreciated.

Thanks in advance!

[chaofengc]

GMAD serves a different purpose from standard IQA evaluation, so it’s not currently included in the repo. That said, it would be nice to have something like a standalone module (e.g., pyiqa/gmad_test.py) to support this use case.

For now, I’d recommend implementing GMAD separately and using IQA-pytorch models for inference within your own evaluation loop. A simple gmad_test.py could wrap model predictions and handle the pair selection / disagreement optimization logic externally.