There is no problem running on the isic2017 dataset, but there will be an error when running on one's own dataset. How can this be solved?

OSUPCVLab / MobileUNETR

Official Implementation of MobileUNETR: A Lightweight End-To-End Hybrid Vision Transformer For Efficient Medical Image Segmentation (ECCV2024) (Oral)

39 stars 4 forks source link

There is no problem running on the isic2017 dataset, but there will be an error when running on one's own dataset. How can this be solved? #5

Open xiachashuanghua opened 1 month ago

xiachashuanghua commented 1 month ago

tpwxtd commented 3 weeks ago

Could you teach me how to reproduce this code? I didn't quite understand how the data is processed and where it needs to be placed. If you could teach me, I would be very grateful.

srperera commented 1 week ago

hey sorry for the delayed response, got caught up with a lot of work. and I think the issue is a simple fix. the shapes should be [B, C, H, W] however you have [8, 1, 512, 512, 3] there's an extra dimension being added there. if this is persisting, maybe perform a .squeeze() operation on the tensor to fix the shape issues. It's probably happening inside the data loader where you have an extra dimension.