uncbiag/SegNext

NoC (Number of Clicks) for SAM and SAM-HQ

Closed this issue · 4 comments

how the NoC (Number of Clicks) for SAM and SAM-HQ is tested, is the get_next_point() function used during testing the same as in segnext?
or any existing code available?

thanks a lot

Hi, Thanks for your wonderful work! I also run the eval_sam scripts, my results is below
NoC@90 = 5.147826086956521
NoC@95 = 10.785507246376811
NoF@95 = 150
5-mIoU = 0.9091643193449243,
why there is a little difference

@qinliuliuqin The results may differ slightly given different versions of PyTorch and other packages. Did you use python=3.10, torch=2.2.0, torchvision=0.17.0?