scipy.misc.imsave is deprecated

[rien333]

According to the scipy documentation, the scipy.misc.imsave function is deprecated. More so, it has been completely removed in newer scipy 1.3+ versions, so that nii2png now fails:

$ python nii2png.py -i in.png -o foo/
Reading NIfTI file...
Saving image...
Traceback (most recent call last):
  File "nii2png.py", line 158, in <module>
    main(sys.argv[1:])
  File "nii2png.py", line 142, in main
    scipy.misc.imsave(image_name, data)
AttributeError: module 'scipy.misc' has no attribute 'imsave'

However, the documentation linked above does recommend another way to save images.

[alexlaurence]

Thanks for this, this is something I have been trying to overcome. Any contributions are welcome on this issue... for the time being, I recommend downgrading scipy for nii2png

[rien333]

What about just doing what the documentation says, and introducing a new dependency to save images? I think that's what I did to fix this, and it worked just fine from what I can remember.

[petremihaivalentin]

What about just doing what the documentation says, and introducing a new dependency to save images? I think that's what I did to fix this, and it worked just fine from what I can remember.

Can confirm. I installed the imageio library and changed all scypi.misc.imsave instances to imageio.imwrite and it worked like a charm.

[aviogit]

imageio.imwrite seems the way to go, I've submitted a PR.

[vcasellesb]

imageio.imwrite seems the way to go, I've submitted a PR.

Hi, I've been trying to get the code to run with imageio.imwrite, but get the following error:

OSError: cannot write mode F as PNG

Does anyone know the fix for this?

Thanks. Best,
Vicent

[OliJimbo]

I've added a PR that fixes this - it's because nibabel deprecated get_data so you have to pass it to numpy.asanyarray

[vcasellesb]

I've added a PR that fixes this - it's because nibabel deprecated get_data so you have to pass it to numpy.asanyarray

Hi,

Thanks for your reply. Wouldn't it be better using nibabel's .get_fdata()? I think that's the recommended way.