Stereo vision concept in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the main purpose of stereo vision in computer vision?

Stereo vision uses two cameras to mimic human eyes. What is the main goal of using stereo vision?

A To improve image brightness

B To estimate the depth of objects in a scene

C To detect edges more accurately

D To increase the color resolution of images

💡 Hint

Think about what having two slightly different views helps us understand about the scene.

Stereo vision helps calculate how far objects are by comparing two images from different viewpoints, similar to how our eyes perceive depth.

❓ Predict Output

intermediate

Output of disparity map calculation code snippet

Given the following Python code using OpenCV to compute a disparity map, what is the shape of the output disparity map?

Computer Vision

import cv2
left_img = cv2.imread('left.png', 0)
right_img = cv2.imread('right.png', 0)
stereo = cv2.StereoBM_create(numDisparities=16, blockSize=15)
disparity = stereo.compute(left_img, right_img)
print(disparity.shape)

A (height, width)

B (width, height)

C (height, width, 3)

D (width, height, 3)

💡 Hint

The disparity map is a single-channel image matching the input image size.

The disparity map has the same height and width as the input images but only one channel, so its shape is (height, width).

❓ Model Choice

advanced

Choosing the best stereo matching algorithm for real-time depth estimation

You want to implement stereo vision on a mobile robot that needs real-time depth estimation. Which stereo matching algorithm is best suited for this task?

A Semi-Global Matching (SGM) algorithm

B Graph Cuts algorithm

C Block Matching (BM) algorithm

D Dynamic Programming algorithm

💡 Hint

Consider the trade-off between speed and accuracy for real-time use.

Block Matching is faster and simpler, suitable for real-time applications, while others are more accurate but slower.

❓ Hyperparameter

advanced

Effect of increasing numDisparities in stereo matching

In stereo vision, what happens if you increase the numDisparities parameter in the block matching algorithm?

A The disparity search range increases, allowing detection of farther objects but increasing computation time

B The image resolution increases, improving depth accuracy

C The block size increases, improving noise resistance but reducing detail

D The number of cameras needed increases

💡 Hint

numDisparities controls how far the algorithm looks for matching points horizontally.

Increasing numDisparities expands the horizontal search range for matching pixels, which helps detect objects at greater distances but requires more computation.

❓ Metrics

expert

Evaluating stereo vision depth accuracy with error metrics

You have ground truth depth data and predicted depth from stereo vision. Which metric best measures the average absolute difference between predicted and true depth values?

A Peak Signal-to-Noise Ratio (PSNR)

B Root Mean Squared Error (RMSE)

C Structural Similarity Index (SSIM)

D Mean Absolute Error (MAE)

💡 Hint

Look for the metric that averages absolute differences without squaring.

MAE calculates the average absolute difference between predicted and true values, making it a straightforward measure of depth accuracy.