Cv2 Image Error: Error: (-215:assertion Failed) !ssize.empty() In Function 'cv::resize'

I'm making an image classifier with TensorFlow and Python, but I'm having an error reading images with CV2. I'm very new to CV2, and I haven't been able to find anything that suffi

Solution 1:

check if there _DS_Store file in the folder you're trying to iterate in the for loop, as it is not a image file cv2.resize() is not able to resize it.

Solution 2:

You're probably reading in an invalid zero-pixel image. Check your image path since the size of your source image is most likely empty and has not been read in properly. You can check the image's shape to ensure that it has a valid (height, width, channels) dimension like this

image = cv2.imread('noexist.png')
print(image.shape)

If you get an AttributeError error similar to this, it means your image path is invalid/corrupted.

Traceback (most recent call last):
  File ".\test.py", line 4, in <module>
    print(image.shape)
AttributeError: 'NoneType' object has no attribute 'shape'

OpenCV will not throw an exception until you try to do something with this NoneType object. So in the next line when you attempt to resize the image, it throws the error. Your situation is reproducible with this simple example

importcv2image= cv2.imread('noexist.jpg')
resize = cv2.resize(image, (64,64))
cv2.waitKey()

Note if you remove the cv2.resize(), an exception will not be thrown. To fix your problem, you can use cv2.error and a try/except block to catch the exception

import cv2

image = cv2.imread('noexist.jpg')
try:
    resize = cv2.resize(image, (64,64))
except cv2.error as e:
    print('Invalid frame!')
cv2.waitKey()

Invalid frame!

Solution 3:

Your imread() call failed. It returned None.

When you try to imread() a file that doesn't exist, or it's not a proper picture file, imread signals that by returning None (in Python), instead of a proper numpy array.

Now, when you pass this None to resize(), resize() notices and throws this error. It tries to check if the argument is a proper array, and that it's not an empty array. In this case, your argument isn't even an array, but the error is the same.

How to fix this: check if the result of imread() is None. If it is None, reading failed. Just continue the loop on to the next picture file name.

img = cv2.imread(path)
if img isNone:
    continue

by the way: the 3 in the second argument to imread means IMREAD_COLOR | IMREAD_ANYDEPTH

Solution 4:

Based on the error it seems there is some issue in the input. It should be a list of non empty arrays. Please verify the path and images for any corrupt image. Also, check the image formats; if any multi band tiff file is present which cv2.imread is not able to read.

Try reading the images in batches to identify the batch having the image that is causing the problem.

Solution 5:

Before reading the image, check if there is an non-existent image. Try this before reading the images:

ifnotos.path.isfile(train_data):
    continue

path = os.path.join(train_data, i)

img = cv2.imread(path, 3)