Why is CV2 not reading the images?

Question

I have a dataset of training images and test images. What I want to do is feed in the training images and resize them to 150x150 size. Then, depending on the class name of the image file, append a label to the array 'y', which is my array of labels.

However, I get this error message:

OpenCV(4.2.0) /io/opencv/modules/imgproc/src/resize.cpp:4045: error: (-215:Assertion failed) !ssize.empty() in function 'resize'

Relevant Part of my code is as follows:

nrows = 150
ncolumns = 150
channels = 3

def read(imgarray):
    x = []
    y = []

    for image in imgarray:
        try:
            x.append(cv2.resize(cv2.imread(image, cv2.IMREAD_COLOR), (nrows,ncolumns), interpolation=cv2.INTER_CUBIC))
        except Exception as e:
            print(str(e))
        if 'chicken' in image:
            y.append(0)
        elif 'cat' in image:
            y.append(1)
        elif 'scoop' in image:
            y.append(2)

    return x,y
x,y = read(train_images) #train_images is composed of ~5400 images, of mixed sizes and image formats

Please can someone tell me why CV2 isn't 'seeing' the images and how I can get the images to be resized?

edit: an example image name is '../input/train/train/chicken (1438).jpg' and the image shape is (340,594, 3)

I am using a Kaggle kernel where my training images and testing images are stored in a directory called 'input'. Training images are in input/train/train/img.jpg and testing images are in input/test/img2.jpg.

Update: when I tried to display the images in train_images:

for image in imgarray:
        #print(image)
        image = mpimg.imread(image)
        showplot = plt.imshow(image)

I got this error:

<built-in function imread> returned NULL without setting an error

which is odd as this previous code worked perfectly fine, displaying the images:

import matplotlib.image as mpimg
for i in train_images[0:3]:
    img=mpimg.imread(i)
    imgplot = plt.imshow(img)
    plt.show()

Update: When I output the images that cause an error, I get this:

Please check this image, has some issues ../input/train/train/scoop (1360).jpg
OpenCV(4.2.0) /io/opencv/modules/imgproc/src/resize.cpp:4045: error: (-215:Assertion failed) !ssize.empty() in function 'resize'

so it seems that an image that should work doesn't for some reason

This is one of the images that threw an error:

Be careful about using `except Exception` like that, see https://stackoverflow.com/questions/54948548/what-is-wrong-with-using-a-bare-except. — AMC, Mar 28 '20 at 18:24
you rules out the obvious, already? That is, print the name of the images as you process them in the loop and make sure each one of them is a good file. — Daemon Painter, Mar 28 '20 at 18:25
@AMC Even when I didn't use try/except, I still got the error from CV2... — Swalkr, Mar 28 '20 at 18:27
@DaemonPainter an example image name is '../input/train/train/chicken (1438).jpg' and the image shape is (340,594, 3) — Swalkr, Mar 28 '20 at 18:33
@DaemonPainter thank you for your reply, my apologies - I don't understand, what do you mean by your sentence "Your script is in a folder that is parallel to /input, in the end."? — Swalkr, Mar 28 '20 at 18:38
Please see [this question](https://stackoverflow.com/questions/52162004/i-am-having-trouble-with-this-error-215assertion-failed-ssize-empty-in-fu) as well. — Daemon Painter, Mar 28 '20 at 18:42
@DaemonPainter I'm working in a Kaggle kernel, and as far as I know I only have this directory: input/train/train/img.jpg and input/test/img2.jpg. My apologies if I have got this wrong, I am new to using a Kaggle kernel. — Swalkr, Mar 28 '20 at 18:43
@DaemonPainter I've had a look at that question, unfortunately it seems very similar to mine yet mine does not work whereas theirs does. — Swalkr, Mar 28 '20 at 18:45
Not an expert on cv2, but should the resize tuple include the color dimension? (nrows,ncolumns, 3) — Stefan, Mar 28 '20 at 18:58
There is some image in the folder with unsupported format/ corrupted. — Zabir Al Nazi, Mar 28 '20 at 19:06
Try to run in a loop and see where the bug happens, look into that image. — Zabir Al Nazi, Mar 28 '20 at 19:06
@furcifer do you know which image formats are unsupported? I can delete those and see if it helps — Swalkr, Mar 28 '20 at 19:09
It's hard to tell, sometimes there are just corrupted images. — Zabir Al Nazi, Mar 28 '20 at 19:13

Zabir Al Nazi · Answer 1 · 2020-03-29T23:19:36.690

0

Simply, ignore images with issues.

The way you are adding labels, even if some images are missed the labels will be added to the array.

nrows = 150
ncolumns = 150
channels = 3

def read(imgarray):
    x = []
    y = []

    for image in imgarray:
        try:
            im = cv2.resize(cv2.imread(image), (nrows,ncolumns)
            x.append(im)
            print(type(im))
            print(im.shape)
            if 'chicken' in image:
                y.append(0)
            elif 'cat' in image:
                y.append(1)
            elif 'scoop' in image:
                y.append(2)
        except Exception as e:
            print(str(e))
            print(f'Please check this image, has some issues {image}')


    return x,y

x,y = read(train_images) #train_images is composed of ~5400 images, of mixed sizes and ima

edited Mar 29 '20 at 23:19

answered Mar 28 '20 at 19:17

Zabir Al Nazi

10,298
4
33
60

I've tried your code and updated my question, it seems that a lot of images are throwing up errors – Swalkr Mar 28 '20 at 19:28
Can you post the errors? also post the screenshot, the snaps of the images which are throwing errors? – Zabir Al Nazi Mar 28 '20 at 19:36
Can you actually check this image as I said? Is it a valid image? Post a screenshot of the image, if it is not a valid image, there is nothing to do except ignoring those cases. – Zabir Al Nazi Mar 28 '20 at 20:14
It's definitely a valid image, so I'm puzzled by what the problem is that stops it being read – Swalkr Mar 30 '20 at 14:55

Why is CV2 not reading the images?

1 Answers1