About code

[dbsdmlgus50]

Hello authors, thank you for your excellent research.

When will the code be released?

Thank you.

[libraboli]

Thank you for your interest in our work. We have indeed been very busy recently, and the relevant code has not been organized yet. We will try our best to open-source the code within one month. Thank you for your understanding.

[dbsdmlgus50]

Thank you for your answer to your question.

I have a question. Did you not use the mask in image prompt? For example, did you not use the technique of using the mask in the reference image to filter the mask from sam, such as EMD in the matcher?

[libraboli]

For image prompts, we employed an unsupervised Salient Object Detection method to filter out some background noise. However, for the final output of SAM, we did not use any filter techniques and directly utilized the output mask from SAM.

[dbsdmlgus50]

Then, how did you solve the problem of sam generating several masks?

Don't you use sam's automaskgenerator?

[luckybird1994]

We follow PerSAM, and when using SAM for inference, we employ the multimask=false mode.

[dbsdmlgus50]

Thank you for your answer!

If i have any other questions, can I come to ask you again? :)

[dbsdmlgus50]

Sorry, I have one more question.

Does your method use the same method as hungraian hungarian matching the predicted mask when conducting the evaluation?

[luckybird1994]

We do not use hungraian hungarian.

[dbsdmlgus50]

But doesn't Persam's mask generator create several masks and make them a little blurry??