how to use compose of transforms correctly?

[lg2578]

I'm using compose of transforms,and there are some errors. I can't find any tips from the api docs.
for examples:

trans=T.Compose([
    T.LoadImage(image_only=True),
    
    T.AddChannel(),
    T.Spacing(pixdim=(1.50,1.50)),
    T.Rotate90(),
    T.CropForeground(),
    T.Resize(spatial_size=(160,160),mode='bilinear'),
    T.EnsureType(),
   
])

• T.Rotate90() lead to error:ValueError: Axes=(1, 2) out of range for array of ndim=2.

trans=T.Compose([
T.LoadImage(image_only=True),

T.AddChannel(),
T.Spacing(pixdim=(1.50,1.50)),
T.CropForeground(select_fn=lambda x: x>30),
T.Resize(spatial_size=(160,160),mode='bilinear'),
T.EnsureType(),

])

• T.Resize(spatial_size=(160,160),mode='bilinear'), lead to error:Input and output sizes should be greater than 0, but got input (H: 0, W: 1) output (H: 160, W: 160)
• if x>0 in T.CropForeground(select_fn=lambda x: x>30), T.Resize works

are there some principles,I need to know?

[Nic-Ma]

Hi @lg2578 ,

Thanks for your interest here.
May I know your MONAI version?
I think the default axes of Rotate90 is (0, 1) is the latest version.
The select_fn of CropForeground is to define what's "foreground", seems your dataset uses intensity > 30 as foreground?

Thanks.

[lg2578]

my version is 0.8.1。
default settings for CropForeground is x>0 ,and it works，I just want to try how about intensity > 30,and get an error. it seems when intensity > 30 the output size is H:1 W:0,it's not the truth.

[Nic-Ma]

I am afraid your test image may have no intensity > 30, I will test this corner case and get back to you soon.

Thanks.

[Nic-Ma]

Hi @lg2578 ,

I tried to execute some test programs locally, didn't see any similar issue.
Could you please help share the value range of your image? Especially the pixel map of intensity > 30?

Thanks.