Closed 1320396502 closed 3 months ago
These NAN values are caused by the PIE-BECH, since for some images the editing needs to be applied to the entire image so there is not "unedited part" to calculate the metrics. This is a normal behavior and you can just calculate the average for the whole dataset. Thanks!
What to do with a lot of NAN values when re-testing metrics, I'd like to ask the author how handled it,Thanks a lot!