OpenCV python inRange method causes error in C++ module

Question

I am getting the following error when I call my method [given after the error].

error: /feedstock_root/build_artefacts/opencv_1496434080029/work/opencv-3.2.0/modules/core/src/arithm.cpp:1984: error: (-215) lb.type() == ub.type() in function inRange

The code that produces the code is right here. I don't see anything wrong with the code I feed to the function cv2.inRange

def red_filter(rgb_image):
    hsv_image = cv2.cvtColor(rgb_image, cv2.COLOR_RGB2HSV)

    avg_brightness = average_brightness(hsv_image)
    avg_saturation = average_saturation(hsv_image)

    # Define our color selection boundaries in HSV values
    lower_red = np.array([0, avg_saturation, avg_brightness]) 
    upper_red = np.array([20,255,255])

    # Define the masked area
    hsv_mask = cv2.inRange(hsv_image, lower_red, upper_red)

    # Copy image
    hsv_masked_image = np.copy(hsv_image)

    #  Mask the image to let the light show through
    hsv_masked_image[hsv_mask != 0] = [0, 0, 0]

    # Display it!
    plt.imshow(hsv_masked_image)

Any idea what the issue may be? I couldn't find the solution in other related questions: Question 1 & Question 2

my guess is that `avg_saturation` and `avg_brightness` are probably floats and not int... thus the lower bound (lb) and upper bound(ub) have different types — api55, Mar 04 '18 at 19:58
@api55 Would you like to put that in an answer? It works when I floor the values. — heretoinfinity, Mar 04 '18 at 21:10

score 2 · Accepted Answer · answered Mar 05 '18 at 12:34

Lets start by explaining the error:

lb.type() == ub.type() in function inRange

This means that it failed in the assert that checks if the lower bound (lb) and upper bound (up) are of the same type in the function inRange.

Looking at your code, the upper bound looks like it is int numbers:

upper_red = np.array([20,255,255])
print (upper_red.dtype) # this prints dtype('int32')

Now, the lower bound has 2 variables, which I do not know what are they (float, int, etc). I will assume they are floats, lets see what happens if I put two float numbers.

lower_red  = np.array([0, 1.2, 2.7])
print (lower_red .dtype) # this prints out dtype('float64')

As you can see they are not the same type. Now that the problem is explained, lets continue to the possible solutions:

Easiest one, if you want to have it truncated:

lower_red  = np.array([0, 1.2, 2.7], dtype=np.int32)
print (lower_red.dtype) # this prints out dtype('int32')
print (lower_red) # [0, 1, 2]

This generates the same result as:

lower_red  = np.array([0, int(1.2), int(2.7)])

If you do not want to truncated you can always do round or ceil (floor is the same as truncating).

For example:

avg_saturation = int(np.round(average_saturation(hsv_image)))

or if it is non negative:

avg_saturation = int( average_saturation(hsv_image) + 0.5 )

OpenCV python inRange method causes error in C++ module

1 Answers1