First image processing program in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the output of this image processing code?

Consider this Python code that loads an image, converts it to grayscale, and prints the shape of the resulting image array. What will be printed?

Computer Vision

import cv2
image = cv2.imread('sample.jpg')
gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
print(gray_image.shape)

A (width, height)

B (height, width, 3)

C Error: cv2.cvtColor requires 3-channel input

D (height, width)

💡 Hint

Grayscale images have only one channel, so the shape has two dimensions.

The cv2.cvtColor function converts the color image to grayscale, resulting in a 2D array with shape (height, width).

❓ Model Choice

intermediate

Which model is best for simple image classification?

You want to classify images of cats and dogs with a small dataset. Which model is the best choice?

A A decision tree using image file names as features

B A simple logistic regression model using raw pixel values

C A pretrained convolutional neural network with transfer learning

D A deep convolutional neural network with many layers trained from scratch

💡 Hint

Using pretrained models helps when data is limited.

Pretrained CNNs with transfer learning leverage learned features and perform well on small datasets.

❓ Hyperparameter

advanced

Which hyperparameter change will most reduce overfitting in image classification?

You trained a CNN that performs well on training images but poorly on new images. Which hyperparameter adjustment will most likely reduce overfitting?

A Add dropout layers with a dropout rate of 0.5

B Increase the number of convolutional layers

C Decrease the batch size to 1

D Remove data augmentation

💡 Hint

Dropout randomly disables neurons during training to prevent memorization.

Adding dropout helps the model generalize better by reducing reliance on specific neurons.

❓ Metrics

advanced

Which metric is best to evaluate image segmentation quality?

You built a model to segment objects in images. Which metric best measures how well the predicted segments match the true segments?

A Accuracy

B Intersection over Union (IoU)

C Mean Squared Error

D Precision

💡 Hint

This metric compares the overlap between predicted and true segments.

IoU measures the overlap area divided by the union area of predicted and true segments, making it ideal for segmentation.

🔧 Debug

expert

What error does this image processing code raise?

This code tries to resize an image but raises an error. What is the error?

Computer Vision

import cv2
image = cv2.imread('nonexistent.jpg')
resized = cv2.resize(image, (100, 100))

A cv2.error: OpenCV(…) error: (-215:Assertion failed) !ssize.empty() in function 'resize'

B FileNotFoundError: No such file or directory: 'nonexistent.jpg'

C TypeError: 'NoneType' object is not subscriptable

D ValueError: invalid literal for int() with base 10

💡 Hint

cv2.imread returns None if the file is missing, causing resize to fail.

cv2.resize fails because the input image is None, triggering an assertion error inside OpenCV.