Python image_list to np.array

Question

I used python's list to add multiple numpy.array images read by opencv:

[array([[[167, 145, 121],
        [164, 142, 118],
        [167, 145, 121],
        ...,          
        [248, 243, 214],
        [246, 242, 213],
        [249, 245, 216]],
       [[172, 150, 126],
        [168, 146, 122],
        [163, 141, 117],
        ...,
        [249, 244, 214],
        [246, 242, 213],
        [248, 244, 215]],
       ...,]

I want to turn the outermost list into a numpy array, that is, a 4-axis tensor np.array:

array([[[[167, 145, 121],
        [164, 142, 118],
        [167, 145, 121],
        ...,
        [248, 243, 214],
        [246, 242, 213],
        [249, 245, 216]],

       [[168, 146, 122],
        [164, 142, 118],
        [164, 142, 118],
        ...,
        [248, 243, 214],
        [246, 242, 213],
        [249, 245, 216]],

       [[172, 150, 126],
        [168, 146, 122],
        [163, 141, 117],
        ...,
        [249, 244, 214],
        [246, 242, 213],
        [248, 244, 215]],

       ...,]

However, if I use np.array(mylist) directly, it becomes:

array([array([[[167, 145, 121],
        [164, 142, 118],
        [167, 145, 121],
        ...,
        [248, 243, 214],
        [246, 242, 213],
        [249, 245, 216]],

       [[168, 146, 122],
        [164, 142, 118],
        [164, 142, 118],
        ...,
        [248, 243, 214],
        [246, 242, 213],
        [249, 245, 216]],
        ...,
        [249, 244, 214],
        [246, 242, 213],
        [248, 244, 215]],
        ....]

Is there a way to convert this?

Gabriel A. · Accepted Answer · 2021-04-07 17:18:10Z

Does all the images have the same shape (width, heigh and number of channels)? If so, doing a np.array(mylist) should have worked just fine. For example, here I created 10 random images:

my_list = [np.random.randint(0, 255, size=(1920, 1080, 3), dtype=np.uint8) for i in range(10)]
converted = np.array(my_list)

Which results in what you expects:

 array([[[[213,  60,  51],
         [229, 125, 207],
         [104, 139, 243],
         ...,
         [166, 219,  32],
         [116,  27, 108],
         [ 99,  79,  21]],

        [[176, 141, 170],
         [107, 131,  83],
         [ 23, 210, 126],
         ...,
         [147,  41, 167],
         [203, 118,  86],
         [175,   5,  88]]]], dtype=uint8)

Now, if there are images with different shapes, you need to manually define the resulting shape. Otherwise it will fail and give you a warning (VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes is deprecated.)

For instance, I created a random list of images in different sizes and selected the biggest dimension, padding the results with zeros.

my_list = [np.random.randint(0, 255, size=(1920-i, 1080-i, 3), dtype=np.uint8) for i in range(10)]
largest_shape = np.max(np.array([m.shape for m in my_list]), axis=0)
result = np.zeros([len(my_list)]+largest_shape.tolist())
for i, m in enumerate(my_list):
    result[i, :m.shape[0], :m.shape[1], :m.shape[2]] = m

Collectives™ on Stack Overflow

Python image_list to np.array

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related