Question about the “Interactive Seg” metric

[Heyy7777]

Hi, thanks for sharing your great work!
I have a question regarding the Interactive Seg. results reported in the paper.

Could you please clarify how the Interactive Seg. score is computed?
Is it obtained by averaging (or taking the max of) the three segmentation results?

Also, could you please confirm whether the released P3SAM model is the same “full” version used in the paper?

Thanks a lot for your time and help!

[JZ-Wu]

Same issue. The metrics of the model seem abnormal. When I was trying to reproduce this work, it feels like this paper didn't take the average IoU as it's result(max IoU instead, which can almost exactly match). Could you please explain it so that I can be sure if there is anything wrong with my trial?