Thanks for your impressive work in medical image segmentation. I would like to ask how many feature sizes you use in your UNETR paper. Looking forward to your reply.
Hi @Jason7282 , the default is 16 for UNETR and for the paper. It forms a moderate model complexity, you can also try larger (e.g., 32) for better performance. It can depend on different tasks.
Thanks.
Dear Authors,
Thanks for your impressive work in medical image segmentation. I would like to ask how many feature sizes you use in your UNETR paper. Looking forward to your reply.