Mode incorrectly determined on Image.fromarray with bool input dtype

[akhmerov]

What did you do?

I observed that Image.fromarray works incorrectly on bool array input, likely due to an incorrectly determined mode or rawmode.

import PIL
import numpy as np
print(np.array(PIL.Image.fromarray(np.eye(5, dtype=bool))))

What did you expect to happen?

Same as print(np.eye(5, dtype=bool)), so:

[[ True False False False False]
 [False  True False False False]
 [False False  True False False]
 [False False False  True False]
 [False False False False  True]]

What actually happened?

[[ True False False False False]
 [False False False False False]
 [False False False False  True]
 [False False False  True False]
 [False False  True False False]]

What versions of Pillow and Python are you using?

Pillow: 5.1.0
python: '3.6.5 | packaged by conda-forge | (default, Apr 6 2018, 13:39:56) \n[GCC 4.8.2 20140120 (Red Hat 4.8.2-15)]'
numpy: 1.14.2

Workaround

Use Image.fromarray(boolean_array * np.uint(255))

[aclark4life]

@akhmerov How about a PR to address this issue?

[akhmerov]

@aclark4life sorry, I'm not familiar enough with the codebase and especially with the low level numpy format to identify the source of the issue and fix it efficiently. (I did check if I could come up with an easy fix when trying to figure out what is happening.)

[wiredfool]

Probably related to: #2044 and #350

[radarhere]

I've created PR #3790 to resolve this.

[nico-zck]

Bad news for guys, this bug still exists:

print(np.array(Image.fromarray(np.eye(5, dtype=bool))))
[[ True False False False False]
 [False False False False False]
 [False False False False  True]
 [False False False  True False]
 [False False  True False False]]

Even worse, it will crash when you set argument mode='1' for fromarray function:

print(np.array(Image.fromarray(np.eye(5, dtype=bool), mode='1')))
[[False False False False False]
 [False False False False False]
 [False False False False False]
 [False False False False False]
 [False False False False False]]

Related issues:
#350
#1228
#1727
Oops, this bug has a long history since 2013.

My environment:
python:3.7.3
Pillow: 6.0
numpy: 1.16.3

[hugovk]

@Zhangchenkai The fix is not in Pillow 6.0.0 (released ~2019-04-01), it will be in Pillow 6.1.0 due for release on 2019-07-01.

Please retest with master or wait for the release.

[stranskj]

The bug is still in releases on Pypi. Pillow 7.1.1, Python 3.8.2

[akhmerov]

@stranskj can't confirm with the same versions of Pillow and Python.

EDIT: just in case, numpy version 1.18.1, and I have a conda-forge + linux installation.

[hugovk]

Also cannot reproduce (both installed from PyPI via pip):

Python 3.8.2 (v3.8.2:7b3ab5921f, Feb 24 2020, 17:52:18)
[Clang 6.0 (clang-600.0.57)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(.startup.py)
(imported collections, datetime, itertools, math, os, pprint, re, sys, time)
>>> from PIL import Image
>>> import numpy as np
>>> Image.__version__
'7.1.1'
>>> np.__version__
'1.18.2'
>>> print(np.array(Image.fromarray(np.eye(5, dtype=bool))))
[[ True False False False False]
 [False  True False False False]
 [False False  True False False]
 [False False False  True False]
 [False False False False  True]]
>>> print(np.eye(5, dtype=bool))
[[ True False False False False]
 [False  True False False False]
 [False False  True False False]
 [False False False  True False]
 [False False False False  True]]
>>> np.array(Image.fromarray(np.eye(5, dtype=bool))) == np.eye(5, dtype=bool)
array([[ True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True]])

@stranskj If you still have a problem, please open a new issue and fill in the template with more details. Thanks.

https://github.com/python-pillow/Pillow/issues/new/choose