Error when 'cv.dnn.blobFromImage'

[Zhang__Runze]

Hi,

I tried to run the pre-trained model 'license-plate-recognition-barrier-0001'. But got error like:

$ python3 lr.py

6768

(1, 3, 24, 94)

Traceback (most recent call last):

  File "lr.py", line 19, in <module>

    blob = cv.dnn.blobFromImage(frame, size=(24, 94), ddepth=cv.CV_8U)

cv2.error: OpenCV(4.1.1-openvino) /home/jenkins/workspace/OpenCV/OpenVINO/build/opencv/modules/imgproc/src/resize.cpp:3363: error: (-215:Assertion failed) !dsize.empty() in function 'resize'

My python script:

import cv2 as cv

import numpy as np

# Load the model.

net = cv.dnn.readNet('./FP16/license-plate-recognition-barrier-0001.xml',

                     './FP16/license-plate-recognition-barrier-0001.bin')

# Specify target device.

net.setPreferableTarget(cv.dnn.DNN_TARGET_MYRIAD)

# Read an image.

image = cv.imread('prepimg.jpg')

image = cv.resize(image, (94, 24))

cv.imwrite('mid-image.jpg',image)

image = image[np.newaxis, :, :, :]     # Batch size axis add

image = image.transpose((0, 3, 1, 2))

print('image.size',image.size)

print('image.shape',image.shape)

if image is None:

    raise Exception('Image not found!')

# Prepare input blob and perform an inference.

blob = cv.dnn.blobFromImage(image, size=(24, 94), ddepth=cv.CV_8U)

net.setInput(blob)

out = net.forward()

The python script was modified from the example. And form the log, I am sure that the image file was loaded, and the shape should be right. I just cannot find why this error happened. Have someone saw this error before?

Thanks,

Runze

[Cary_P_Intel1]

Hi, Zhang Runze,

Have you tried the blob = cv.dnn.blobFromImage(image, size=(94, 24), ddepth=cv.CV_8U)? Since for the blobFromImage it accepts the size of (Width, Height). From you original data, looked like your model width is 94 and height is 24.

You can refer to the "preparing input" section in the link below:

https://www.arunponnusamy.com/yolo-object-detection-opencv-python.html ;

[Zhang__Runze]

Hi Cary P, thanks for helping!

But the error is same after I changed the code.

python3 lr.py

image.size 6768

image.shape (1, 3, 24, 94)

Traceback (most recent call last):

  File "lr.py", line 20, in <module>

    blob = cv.dnn.blobFromImage(image, size=(94, 24), ddepth=cv.CV_8U)

cv2.error: OpenCV(4.1.1-openvino) /home/jenkins/workspace/OpenCV/OpenVINO/build/opencv/modules/imgproc/src/resize.cpp:3363: error: (-215:Assertion failed) !dsize.empty() in function 'resize'

Cary P. (Intel) wrote:

Hi, Zhang Runze,

Have you tried the blob = cv.dnn.blobFromImage(image, size=(94, 24), ddepth=cv.CV_8U)? Since for the blobFromImage it accepts the size of (Width, Height). From you original data, looked like your model width is 94 and height is 24.

You can refer to the "preparing input" section in the link below:

https://www.arunponnusamy.com/yolo-object-detection-opencv-python.html 

[Shubha_R_Intel]

Dear Zhang, Runze

Are you sure your image is valid ? Please study the following issue ticket. Also OpenVino 2019R3 was just released. Please upgrade if you haven't already.

https://github.com/the-house-of-black-and-white/opencv-dnn-demo/issues/1

Thanks,

Shubha