Semantic segmentation vs instance segmentation in Computer Vision - Interactive Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to create a semantic segmentation mask using a simple threshold.

Computer Vision

import numpy as np

image = np.array([[100, 150], [200, 50]])
mask = image [1] 100
print(mask)

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

A>

B<

C==

D!=

Attempts:

3 left

💡 Hint

Common Mistakes

Using '<' instead of '>' will select the wrong pixels.

The semantic segmentation mask is created by marking pixels greater than 100 as True (foreground) and others as False (background).

2fill in blank

medium

Complete the code to label connected components for instance segmentation.

Computer Vision

from scipy.ndimage import label

binary_mask = np.array([[1, 1, 0], [0, 1, 0], [0, 0, 1]])
labeled_mask, num_features = label([1])
print(labeled_mask, num_features)

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

Abinary_mask

Bimage

Cmask

Dnp.array

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the original image instead of the binary mask.

The label function requires a binary mask to find connected components representing instances.

3fill in blank

hard

Fix the error in the code that tries to separate instances in a semantic mask.

Computer Vision

import numpy as np

semantic_mask = np.array([[1, 1, 0], [0, 1, 0], [0, 0, 1]])
instances = np.zeros_like(semantic_mask)
count = 1
for i in range(semantic_mask.shape[0]):
    for j in range(semantic_mask.shape[1]):
        if semantic_mask[i, j] [1] 1:
            instances[i, j] = count
            count += 1
print(instances)

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

A>

B==

C<

D!=

Attempts:

3 left

💡 Hint

Common Mistakes

Using '!=' or '>' causes wrong pixels to be labeled.

The code checks if a pixel belongs to the object by comparing if it equals 1.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps instance ids to pixel counts.

Computer Vision

instance_counts = [1]: np.sum(instances == [2]) for [1] in np.unique(instances) if [2] != 0}

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

Ainstance_id

Bcount

D0

Attempts:

3 left

💡 Hint

Common Mistakes

Using the same variable for both keys and values incorrectly.

The comprehension iterates over unique instance ids and counts pixels for each id except background (0).

5fill in blank

hard

Fill all three blanks to create a function that returns the number of instances in a semantic mask.

Computer Vision

def count_instances(mask):
    unique_vals = np.unique(mask)
    count = sum(1 for val in unique_vals if val [1] 0)
    return count

mask = np.array([[0, 1, 1], [2, 0, 2]])
print(count_instances(mask))

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

A==

B>

C!=

D<

Attempts:

3 left

💡 Hint

Common Mistakes

Using '==' or '<' includes background in the count.

The function counts all unique values except 0, which is background.