Closed prakashjayy closed 3 months ago
I have edited the code on my local such that window_size is tuple instead of int. since the input size of shape (Dp, Hp, Wp). the output shape also should be of the same. The following line should be re-written as
https://github.com/uni-medical/SAM-Med3D/blob/a5227965b78b5cf74010df32877b96d21f5b4b45/segment_anything/modeling/image_encoder3D.py#L323
x = x.permute(0, 1, 4, 2, 5, 3, 6, 7).contiguous().view(B, Hp, Wp, Dp, -1)
to
x = x.permute(0, 1, 4, 2, 5, 3, 6, 7).contiguous().view(B, Dp, Hp, Wp, -1)
Thx for the bug report. Will fix it in the next patch.
I have edited the code on my local such that window_size is tuple instead of int. since the input size of shape (Dp, Hp, Wp). the output shape also should be of the same. The following line should be re-written as
https://github.com/uni-medical/SAM-Med3D/blob/a5227965b78b5cf74010df32877b96d21f5b4b45/segment_anything/modeling/image_encoder3D.py#L323
to