Table extraction from images in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load an image using OpenCV for table extraction.

Computer Vision

import cv2
image = cv2.[1]('table_image.png')

Drag options to blanks, or click blank then click option'

Aimshow

Bimread

Cimwrite

Dresize

Attempts:

3 left

💡 Hint

Common Mistakes

Using cv2.imshow() instead of cv2.imread() to load the image.

Trying to use cv2.imwrite() which saves an image, not loads it.

The function cv2.imread() loads an image from a file into memory, which is the first step in processing an image for table extraction.

2fill in blank

medium

Complete the code to convert the loaded image to grayscale for easier table detection.

Computer Vision

gray_image = cv2.cvtColor(image, [1])

Drag options to blanks, or click blank then click option'

Acv2.COLOR_BGR2GRAY

Bcv2.COLOR_BGR2RGB

Ccv2.COLOR_GRAY2BGR

Dcv2.COLOR_RGB2BGR

Attempts:

3 left

💡 Hint

Common Mistakes

Using cv2.COLOR_BGR2RGB which changes color space but not to grayscale.

Using cv2.COLOR_GRAY2BGR which converts grayscale to color, opposite of what is needed.

Converting the image to grayscale using cv2.COLOR_BGR2GRAY simplifies the image and helps in detecting table structures.

3fill in blank

hard

Fix the error in the code to detect edges using Canny edge detection for table boundary detection.

Computer Vision

edges = cv2.Canny(gray_image, [1], 150)

Drag options to blanks, or click blank then click option'

A50

B200

C300

D10

Attempts:

3 left

💡 Hint

Common Mistakes

Setting the first threshold higher than the second, which causes poor edge detection.

Using very low values that cause too many edges.

The first threshold in Canny edge detection is usually lower than the second. Using 50 and 150 is a common pair for detecting edges effectively.

4fill in blank

hard

Fill both blanks to apply dilation and find contours for table structure extraction.

Computer Vision

kernel = cv2.getStructuringElement(cv2.MORPH_RECT, ([1], [2]))
dilated = cv2.dilate(edges, kernel, iterations=1)

Drag options to blanks, or click blank then click option'

A15

B5

C10

D20

Attempts:

3 left

💡 Hint

Common Mistakes

Using very large kernel sizes that over-dilate and merge unrelated edges.

Using kernel sizes that are too small to connect table lines.

A rectangular kernel of size (5, 10) is commonly used to dilate edges vertically and horizontally to connect table lines.

5fill in blank

hard

Fill all three blanks to extract bounding boxes of detected table cells from contours.

Computer Vision

contours, _ = cv2.findContours(dilated, cv2.RETR_EXTERNAL, [1])
bounding_boxes = [cv2.boundingRect(c) for c in contours if cv2.contourArea(c) > [2]]
sorted_boxes = sorted(bounding_boxes, key=lambda b: (b[[3]], b[0]))

Drag options to blanks, or click blank then click option'

Acv2.CHAIN_APPROX_SIMPLE

B1000

C1

Dcv2.CHAIN_APPROX_NONE

Attempts:

3 left

💡 Hint

Common Mistakes

Using CHAIN_APPROX_NONE which is slower and unnecessary here.

Not filtering small contours causing noise in bounding boxes.

Sorting by x before y which disrupts reading order.

cv2.CHAIN_APPROX_SIMPLE compresses contour points, 1000 filters out small contours, and sorting by b[1] (y-coordinate) then b[0] (x-coordinate) orders boxes top to bottom, left to right.