SegNext require many clicks when I run demo on DAVIS test image

[Z-MU-Z]

00013.jpg the checkpoint is vitb_sa2_cocolvis_hq44k_epoch_0.pth

I compare with simpleclick, which only require a few clicks

another case 00340.jpg

[qinliuliuqin]

Thank you for pointing that out. As we demonstrated in Table 1, SimpleClick actually outperforms SegNext in terms of NoC95 and NoF95. Since SegNext is still in the research stage and requires further refinement, it may not be the best choice for real-world applications.

[Z-MU-Z]

Thanks for your reply. BTW, is this demo support box input?

[qinliuliuqin]

This demo is derived from RITM and currently supports only points. I'm developing a more general GUI for diverse prompts, which will be released in a follow-up.

[lmr2706]

Hello, are you saying that your current demo program can only implement the "point" prompt type? But I noticed that the paper involves the collaboration of multiple prompt types—does the code you provided cover this part?

[qinliuliuqin]

@lmr2706 Hi, please refer to https://github.com/uncbiag/SegNext/issues/12.