BubbleGrader

BubbleGrader is a tool to analyse and detect the bubbles filled in OMR answer sheets and accordingly grade the answer sheet.

What is OMR?

OMR stands for Optical Mark Recognition is the process of automatically analyzing human-marked documents and interpreting their results.

Tools and libraries used

• opencv
• numpy
• imutils

To install all these dependencies, use python -m pip install requirements.txt.
Requirements.txt is provided with this repository.

Steps

1. Detect the exam paper from the image
2. Apply a perspective transform to extract the top-down, birds eye view of the exam sheet.
3. Extract the set of bubbles from the perspective transformed exam sheet.
4. Sort the bubbles from the exam sheet into rows. 1 row corresponding to 1 question.
5. Determine the marked bubbles from each row.
6. Cross-check the marked answer with the answer key.
7. Repeat this process for all rows.

Implementation

The line 16 from the file test_grader.py defines the answer key for our exam.

ANSWER_KEY = { 0:1, 1:4, 2:0, 3:3, 4:1 }

It is basically a dictionary.
Right from line 18 to 25 we preprocess the image.

if args["crop"] == 1:
    image = imutils.resize(cv2.imread(args["image"]), width=700)
else:
    image = cv2.imread(args["image"])
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, (5,5), 0)
edged = cv2.Canny(blurred, 75, 200)

By preprocessing, I mean, resizing, blurring, converting the image to grayscale and edge detection.
Each of these processes help reduce the complexity of the further operations to be done on the image.

The next step is to identify the exam sheet present in the image. Here, we use contours to find the exam sheet.

cnts = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)

After this, the cnts variable stores a list of contours. Now, in order to identify the exam paper, (which will be a rectangle) we loop through the contours to find a contour having 4 corner points. This contour will correspond to the exam sheet. The code to achieve this is

for cnt in cnts:
        perimeter = cv2.arcLength(cnt, True)
        approx = cv2.approxPolyDP(cnt, 0.02*perimeter, True) #approximating the contour to 2% of the found contour.

        if len(approx) == 4:
            docCnt = approx
            break

The function cv2.approxPolyDP approximates the contour cnt to 2% of the original contour. For more info about this function, visit here.
Variable docCnt now has the contour of the exam paper.

Now, we apply perspective transform on the image to get a birds eye view.

warped = four_point_transform(gray, docCnt.reshape((4, 2)))

The birds eye view of the exam sheet is:

After this, we threshold the image using Binary / Otsu threshold

Binarization again allows us to find contours from the image. This time, the contours will be the bubbles from the image. We append all such contours in a list called questionContours

for c in cnts:
    (x, y, w, h) = cv2.boundingRect(c)
    aspect_ratio = w / float(h)

    if w >= 20 and h >= 20 and aspect_ratio >= 0.9 and aspect_ratio <= 1.1:
	    questionContours.append(c)

Here, we calculate the bounding box around each contour and check whether it is actually a bubble by checking aspect ratio and width-height. After validations, we append the contour to questionContours.

for (q, i) in enumerate(np.arange(0, len(question_cnts), 5)):
    cnts = contours.sort_contours(question_cnts[i: i+5])[0]
    bubbled = None

    for (j, c) in enumerate(cnts):
        mask = np.zeros(thresh.shape, dtype='uint8')
        cv2.drawContours(mask, [c], -1, 255, -1)

        mask = cv2.bitwise_and(thresh, thresh, mask=mask)
        total = cv2.countNonZero(mask)

        if bubbled is None or bubbled[0] < total:
            bubbled = (total, j)

This arranges the contours in rows, each row corresponding to each question and identifies the filled bubble. total = cv2.countNonZero(mask) counts number of non zero pixels in the contour. After this, we compare the filled bubble index to that of the correct answer in our ANSWER_KEY and print the corresponding result and draw the appropriate contour on the image.