Resizing images in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the shape of the resized image?

Given the following code that resizes an image using OpenCV, what will be the shape of the output image?

Computer Vision

import cv2
import numpy as np

image = np.zeros((100, 200, 3), dtype=np.uint8)
resized_image = cv2.resize(image, (50, 50))
print(resized_image.shape)

A (50, 50)

B (50, 50, 3)

C (100, 50, 3)

D (50, 100, 3)

💡 Hint

Remember that OpenCV uses (width, height) for resizing and color images keep 3 channels.

The original image has shape (100, 200, 3). The resize function uses (width, height) = (50, 50), so the output shape is (height, width, channels) = (50, 50, 3).

🧠 Conceptual

intermediate

Why resize images before training a model?

Why do we usually resize images to a fixed size before feeding them into a machine learning model?

A To ensure all images have the same size so the model input shape is consistent.

B To reduce the number of images in the dataset.

C To increase the number of colors in the image for better accuracy.

D To convert images from color to grayscale automatically.

💡 Hint

Think about how models expect input data shapes.

Models require inputs of the same shape to process batches efficiently. Resizing ensures all images match the expected input size.

❓ Hyperparameter

advanced

Choosing interpolation method for resizing

Which interpolation method is best suited for enlarging images to preserve quality?

A cv2.INTER_CUBIC

B cv2.INTER_LINEAR

C cv2.INTER_AREA

D cv2.INTER_NEAREST

💡 Hint

Some methods are better for shrinking, others for enlarging.

INTER_CUBIC uses bicubic interpolation which produces smoother results when enlarging images compared to INTER_LINEAR or INTER_NEAREST.

🔧 Debug

advanced

Why does this resizing code raise an error?

What error will this code raise and why? import cv2 import numpy as np image = np.zeros((100, 100, 3), dtype=np.uint8) resized = cv2.resize(image, (0, 0), fx=2, fy=2) print(resized.shape)

A TypeError because size cannot be (0, 0)

B AttributeError because resize does not accept fx and fy

C ValueError because fx and fy cannot be used with size

D No error, output shape will be (200, 200, 3)

💡 Hint

Check OpenCV documentation for resize parameters.

When size is (0, 0), OpenCV uses fx and fy to scale the image. So no error occurs and output shape doubles.

❓ Metrics

expert

Effect of resizing on model accuracy

You train two image classifiers: one with images resized to 64x64, another with 256x256. Which is the most likely outcome?

A The 256x256 model will always have higher accuracy due to more detail.

B The 64x64 model will always have higher accuracy due to faster training.

C The 256x256 model may have higher accuracy but risks overfitting and longer training time.

D Both models will have identical accuracy regardless of image size.

💡 Hint

Consider trade-offs between image detail and model complexity.

Larger images provide more detail but increase model complexity and training time, which can cause overfitting if data is limited.