search

uni-medical / SAM-Med3D

SAM-Med3D: An Efficient General-purpose Promptable Segmentation Model for 3D Volumetric Medical Image
Apache License 2.0
484 stars 64 forks source link

inference 结果有bug #73

Open ssqianji opened 3 months ago

ssqianji commented 3 months ago

训练时候结果dice有0.8，但infernce预测val数据时候都是0.1多 微信图片_20240619104927 就算是使用sam_med3d_turbo.pth，也是这个结果。

RRouhi commented 2 months ago

Hi,

I have the same issue. The Dice score is about 0.89 and the loss is about 1.2, but I get similar output for my segmentation task for the inference. Can someone help to address this?

adrianzzk commented 2 months ago

很早之前我就发现有这个问题了，推测可能是dataloader中某一部分出bug了，感兴趣可以调试一下

ssqianji commented 1 month ago

很早之前我就发现有这个问题了，推测可能是dataloader中某一部分出bug了，感兴趣可以调试一下

但是它使用的是torch的dataloader，具体是指哪一部分？