autozomm and preMask

[jianjun0407]

Thank you for your sharing. I have two questions in the process of using nnInteractive and would like to ask you for advice:
（1） To add premask results, is session.add_initial_seg_interaction() used?

（2）When I used nnInteractive to segment the following targets, autozoom did not play a role and only segmented a part of the target area. Is there any parameter that needs to be set?

[FabianIsensee]

Hey,
if you already have an initial segmentation mask you can indeed set it with add_initial_seg_interaction. Please make sure the segmentation and image have aligned numpy arrays. You need to load both in the same way. I am just mentioning that because you are transposing axes for the segmentation and would need to have done the same for the image. You can set run_prediction=True to do an auto refinement without interactions. Or set it to False to wait for the first interaction
In 2) it seems like the target organ was not properly predicted from your initial interaction, causing AutoZoom to not be triggered. I suggest to use scribbles or lasso over points as this more clearly communicates to nnInteractive what you want and will more reliably trigger the correct response.
Best,
Fabian