EVF-SAM icon indicating copy to clipboard operation
EVF-SAM copied to clipboard

Published 20 hours ago • verified publisher icon hustvl

How to support multi-class segmentation

Open ceclee opened this issue 7 months ago • 2 comments

Thanks for your codes and I have finished the single-class segmentation. Now I want the multi-class segmentation tasks can be supported, where should the modifications be made? I tried to modify the "pred_mask" part in evf-sam2.py, but I'm not sure how to change the output of "p" to support multiple classes. Currently, it's all set to 1. My gt_masks like (batchsize,num_class,w,h). Thank you to the author for patiently reading through this. If you happen to see this, I would greatly appreciate any guidance or advice you could offer.

ceclee avatar Apr 10 '25 12:04 ceclee

EVF-SAM doesn't support multi-class segmentation within one inference. You may consider batch inference.

CoderZhangYx avatar Apr 10 '25 13:04 CoderZhangYx

Thank you for ur quick response! and what do you mean batch inference? "Batch inference" refers to a certain mode or pattern？ Can you provide me with more details about it?

EVF-SAM doesn't support multi-class segmentation within one inference. You may consider batch inference.

ceclee avatar Apr 10 '25 13:04 ceclee