Haar cascade face detection in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Haar cascade features

Which of the following best describes what Haar cascade features detect in an image?

A Patterns of light and dark areas that resemble edges and lines

B Color gradients and textures in the image

C Pixel intensity histograms across the entire image

D Frequency components using Fourier transform

💡 Hint

Think about how Haar features use simple rectangular shapes to find contrasts.

Haar cascade features detect contrasts between light and dark rectangular regions, which correspond to edges and lines in the image. This helps identify facial structures.

❓ Predict Output

intermediate

Output of face detection code snippet

What will be the output of the following Python code snippet using OpenCV's Haar cascade?

Computer Vision

import cv2
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
img = cv2.imread('group_photo.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5)
print(len(faces))

A A boolean indicating if any face was detected

B A list of coordinates of detected faces

C The number of faces detected in the image

D An error because detectMultiScale requires color image

💡 Hint

Check what the variable faces contains and what len() returns.

The detectMultiScale method returns a list of rectangles for detected faces. Using len(faces) gives the count of faces detected.

❓ Model Choice

advanced

Choosing parameters for Haar cascade detection

You want to detect faces in images with varying sizes and some noise. Which parameter setting for detectMultiScale is best to balance detection accuracy and speed?

A <code>scaleFactor=1.05</code>, <code>minNeighbors=3</code>

B <code>scaleFactor=1.1</code>, <code>minNeighbors=5</code>

C <code>scaleFactor=1.5</code>, <code>minNeighbors=10</code>

D <code>scaleFactor=1.3</code>, <code>minNeighbors=1</code>

💡 Hint

Lower scaleFactor means more scales checked, higher minNeighbors means fewer false positives.

A scaleFactor of 1.1 and minNeighbors of 5 is a common balanced choice that detects faces well while reducing false positives and keeping reasonable speed.

❓ Metrics

advanced

Evaluating Haar cascade face detector performance

Which metric is most appropriate to measure how well a Haar cascade face detector finds all faces without missing any?

A Precision

B F1-score

C Accuracy

D Recall

💡 Hint

Recall measures how many actual faces are detected out of all faces present.

Recall measures the proportion of true faces detected. High recall means few faces are missed, which is important for face detection completeness.

🔧 Debug

expert

Debugging no faces detected error

You run this code but it always prints 0 faces detected, even though faces are visible in the image. What is the most likely cause?

import cv2
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
img = cv2.imread('face.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
print(len(faces))

A The cascade file path is incorrect, so the classifier is empty

B The image is not converted to grayscale before detection

C The detectMultiScale method requires color images, not grayscale

D The image file 'face.jpg' is corrupted and cannot be read

💡 Hint

Check if the cascade classifier loaded properly before detection.

If the cascade file path is wrong, the classifier is empty and detects no faces. Using cv2.data.haarcascades ensures correct path.