KeyError with np.uint32 and numpy.uint32

Question

First off the Error I get:

import numpy as np    
dtype_range = {np.bool_: (False, True),
           np.bool8: (False, True),
           np.uint8: (0, 255),
           np.uint16: (0, 65535),
           np.int8: (-128, 127),
           np.int16: (-32768, 32767),
           np.int64: (-2**63, 2**63 - 1),
           np.uint64: (0, 2**64 - 1),
           np.int32: (-2**31, 2**31 - 1),
           np.uint32: (0, 2**32 - 1),
           np.float32: (-1, 1),
           np.float64: (-1, 1)}

dtype_range[image.dtype.type]

>>>KeyError: <type 'numpy.uint32'>

Crazy thing is, when I do

print image.dtype.type
>>><type 'numpy.uint32'>

Now the main problem is, that this happens in a standard lib of skimage.io, in dtype.py to be exact, and I can not change that source code. So I am wondering how or what can I change in passing my argument image to make it work? My guess would be that there is a problem with the namespace, since it is np and numpy? But how could I influence my data to save it as np.uint32 instead of numpy.uint32??

I am using Python 2.7 btw.

Help would be much appreciated.

EDIT:

I casted image to uint8 by

image = image.astype(uint8)

now this did not throw me an Error, even though

print image.astype(uint8).dtype.type
>>><type 'numpy.uint8'>

which you would expect, of course.

This resolved my issue. But if anyone knows what happened here, I would be glad for an answer just to understand the problem. Thanks anyways :)

EDIT:

Well now this is copied from the output of the IPython console:

 image = np.array([35, 37, 39, 36, 34, 31, 33, 32, 32, 33, 31, 33, 30, 34, 36, 37, 36,
   32, 33, 30, 28, 30, 28, 28, 29, 30, 29, 31, 30, 31, 36, 33, 34, 31,
   34, 35, 34, 32, 29, 26, 25, 27, 25, 26, 25, 27, 30, 30, 28, 26, 28,
   30, 32, 34, 36, 36, 36, 32, 36, 37, 34, 34, 35, 33, 33, 30, 33, 36,
   36, 36, 33, 33, 39, 38, 34, 32, 32, 29, 28, 29, 30, 32, 32, 28, 30,
   32, 34, 30, 28, 32, 34, 34, 35, 33, 35, 33, 33, 35, 37, 39], dtype=uint32)

There is something wrong with your copy paste. The `print` statement should print `` and not just `numpy.uint32`. Otherwise your error is that your are indexing your dict with a string rather than with a type. — JohanL, Jun 22 '17 at 14:14
Yes you are right, I had not actually printed it in the source code, but executed it in the Ipython console of spyder, that is why my output looked different. I changed it to the correct form. — Nils, Jun 22 '17 at 14:46

Tonechas · Accepted Answer · 2017-06-29T10:41:04.853

1

I couldn't reproduce your error using Python 2.7.12 (Spyder 3.1.4) and NumPy 1.11.1. When I run this snippet:

import numpy as np

image = np.array([[35, 37, 39, 36, 34, 31], 
                  [33, 32, 32, 33, 31, 33], 
                  [30, 34, 36, 37, 36, 32],
                  [33, 30, 28, 30, 28, 28]], dtype=np.uint32)

dtype_range = {np.bool_: (False, True),
               np.bool8: (False, True),
               np.uint8: (0, 255),
               np.uint16: (0, 65535),
               np.int8: (-128, 127),
               np.int16: (-32768, 32767),
               np.int64: (-2**63, 2**63 - 1),
               np.uint64: (0, 2**64 - 1),
               np.int32: (-2**31, 2**31 - 1),
               np.uint32: (0, 2**32 - 1),
               np.float32: (-1, 1),
               np.float64: (-1, 1)}

print(image.dtype.type)
print(dtype_range[image.dtype.type])

I got this output:

<type 'numpy.uint32'>
(0, 4294967295L)

My guess is that the issue arises when you cast image to uint32 somewhere in the code. Could you share image and show us the full code?

edited Jun 29 '17 at 10:41

answered Jun 28 '17 at 09:39

Tonechas

13,398
16
46
80

I added a part of `image` because the whole thing is quite big, but I checked and the error arises still, also with this small part of `image` – Nils Jun 28 '17 at 14:17
I replaced `dtype=uint32` by `dtype=np.uint32` in your code and got no error – Tonechas Jun 28 '17 at 14:23
Where did you replace it? – Nils Jun 29 '17 at 09:16
I've just edited my answer. Can you confirm whether you get an error when running the modified code? – Tonechas Jun 29 '17 at 10:44
Well I get my variable `image` from another function so I can not just change the dtype from `uint` to `np.uint` in the script.. Or can I? – Nils Jul 06 '17 at 10:35
You have different options here: 1) `image.dtype = np.uint32`, 2) `image = np.uint32(image)` – Tonechas Jul 06 '17 at 10:40
It worked thank you very much for your help :) Even though I still do not understand why Python has an issue with the namespace – Nils Jul 07 '17 at 13:10

KeyError with np.uint32 and numpy.uint32

1 Answers1