What's the radius of sharpness in your code corresponds to the F1 score in MoGe-2?

[guangkaixu]

Hi, thanks for this great work. MoGe-2 reports promising F1-score in the paper, but I found that there exist three boundary F1 score in your code:

https://github.com/microsoft/MoGe/blob/286873bdaa16fe05086ac2e081d6bc87b7523118/moge/test/metrics.py#L328-L333

Does the F1-score in MoGe-2 represents the average of "radius1_f1", "radius2_f1", "radius3_f1", or it represents either of them?

[EasternJournalist]

Hi, thanks for your question. We used radius=1 in our final evaluation, following Depth Pro. The other two settings were only for our own exploration.

As a disclaimer: the evaluation code for V2 has not been officially released yet. What you see now comes from some uncleaned commits of V1, and there may still be bugs. We will clean up the code and release an official version soon.