Python: Displaying a binary array as image

Question

I have a mask image which is technically a matrix full of True/False values. I would like to view this as an image. First, I converted it into a binary array with (astype(np.uint8))

print('Part Mask', p['masks'][class_id].astype(np.uint8))

but I still can't view it as an image under Python notebook. OpenCV goes crazy and crashes the kernel.

Does anyone know how to view such a structure as an image on Python notebook without crashing the kernel?

    [[0 0 0 ... 0 0 0]
     [0 0 0 ... 0 0 0]
     [0 0 0 ... 0 0 0]
     ...
     [0 0 0 ... 0 0 0]
     [0 0 0 ... 0 0 0]
     [0 0 0 ... 0 0 0]]

or this would work as well ():

[[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 False False ... False False False]
 [False False False ... False False False]]

Thanks in advance.

EDIT: I cannot copy paste the whole code here but basically I have a prediction tensor p, and p[masks][class_id] is what I want to visualize (showing the mask of each class).

enumerator = 0
# run through the instances
for class_id in p['class_ids']:
    #print('Image:', image) # the original input image
    #print('Mask:', merged_mask) # whole masked image
    print('ID: ', class_names[class_id] + str(enumerator))
    #print('Outline Poses: ', ) # mask boundary coordinates
    #print('Pose:',) # mask center coordinates
    print('Part Mask', p['masks'][class_id].astype(np.uint8)) # how to visualize this as an image?
    print('Confidence: ', p['scores'][class_id])
    print('BB: ', p['rois'][class_id]) # get the BB
    print('--------------------------')
    enumerator = enumerator + 1

P.S: Matplotlib does not work either. This is the kind of image I get when I try to print:

Got another error with that, but perhaps it was because of something else, could you suggest a way with matplotlib? — Schütze
– Schütze, Commented Nov 26, 2018 at 8:36
Please do not even think of pasting the whole code here. Create a minimal reproducible example instead. — Stop harming Monica
– Stop harming Monica, Commented Nov 26, 2018 at 9:30
Not sure if this was a bot answer or not @Goyo. I said myself I cannot paste the code here, you don't need to link me that page. — Schütze
– Schütze, Commented Nov 26, 2018 at 9:36
No, it's not. But you do not need to care, just think whether the advice makes sense or not. You have an example of matplotlib working with sample data. It does not work with your data but we don't know why. Knowing that it comes from a masked prediction tensor does not help. Having an actual hardcoded sample does. — Stop harming Monica
– Stop harming Monica, Commented Nov 26, 2018 at 9:49

Dinari · Accepted Answer · 2018-11-26 11:23:10Z

4

Matplotlib should work for you:

import numpy as np
from matplotlib import pyplot as plt


image = np.eye(10)
binary = image > 0
plt.imshow(binary)
plt.show()

With result:

Edit:
Your image is of shape (510,7), what you got above is exactly what you should expect:

import numpy as np
from matplotlib import pyplot as plt


image = np.eye(510)[:,:7]
binary = image > 0
plt.imshow(binary)
plt.show()

Results:

edited Nov 26, 2018 at 11:23

answered Nov 26, 2018 at 8:37

Dinari

2,56716 silver badges29 bronze badges

Sign up to request clarification or add additional context in comments.

9 Comments

Dinari Over a year ago

Post your code, This seems like a problem with the notebook your using, the figure does not get the size it should get.

Schütze Over a year ago

I don't think what you created withnp.eye() there is the same with my data though. Mine is (510, 7) <class 'numpy.ndarray'>

Dinari Over a year ago

Probaly not, but binary = image > 0 create a binary array which I am using to plot. The content of the array does not matter, as long as it is Binary.

Stop harming Monica Over a year ago

@Schütze your image looks ok to me for an array of that shape. What else do you expect?

Dinari Over a year ago

You have an image of (510,7), it will look like a very narrow rectangle, I don't know if it is messed up, as we do not know what is the expected result.

|

Mario · Accepted Answer · 2021-05-31 15:52:49Z

0

Maybe you may use PIL

from PIL import Image

data = [...]

width = len(data[0])
height = len(data)

output_image = Image.new(mode='1', size=(width, height))

for x in range(height):
    for y in range(width):
        pixel_value = data[x][y]
        output_image.putpixel((x,y), pixel_value)


output_image

Output is such this image -><- Click and zoom the image to pixels using the following data:

data = [[0,0,0,0], [0,0,1,1], [1,1,1,0], [0,1,1,0]]

edited May 31, 2021 at 15:52

Mario

2,0844 gold badges34 silver badges81 bronze badges

answered Nov 26, 2018 at 8:40

Crazy

3242 silver badges8 bronze badges

4 Comments

Schütze Over a year ago

Gave an error at output_image.putpixel((x,y), pixel_value) saying SystemError: new style getargs format but argument is not a tuple

Crazy Over a year ago

What version of PIL you use ?

Schütze Over a year ago

Version I have is 5.1.0

Crazy Over a year ago

I have version 5.3.0 and python3.6 , and all works. Try to update PIL version

Collectives™ on Stack Overflow

Python: Displaying a binary array as image

2 Answers 2

9 Comments

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

9 Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related