On some platforms this forces images to be of dtype np.float32 and on others they must be np.float64 because np.float is just an alias for the native floating type on the system. This library should assert floating type input, but it should not care if it is a 32 or 64 bit float.
The fix is to replace all code of the previous style with:
assert img_in.dtype.kind == 'f'
with will work with both float32 and float64 types.
In at least one place this package checks that
On some platforms this forces images to be of dtype
np.float32
and on others they must benp.float64
becausenp.float
is just an alias for the native floating type on the system. This library should assert floating type input, but it should not care if it is a 32 or 64 bit float.The fix is to replace all code of the previous style with:
with will work with both float32 and float64 types.