Bounding box representation in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding bounding box formats

Which of the following bounding box representations uses the coordinates of the top-left corner and the width and height of the box?

A XYWH format

B XYXY format

C Center coordinates with width and height

D Normalized coordinates between 0 and 1

💡 Hint

Think about which format explicitly mentions width and height.

The XYWH format specifies the bounding box by the top-left corner coordinates (X, Y) and the width (W) and height (H) of the box.

❓ Predict Output

intermediate

Output of bounding box conversion code

What is the output of this Python code converting a bounding box from XYWH to XYXY format? box_xywh = [50, 30, 100, 200] box_xyxy = [box_xywh[0], box_xywh[1], box_xywh[0] + box_xywh[2], box_xywh[1] + box_xywh[3]] print(box_xyxy)

Computer Vision

box_xywh = [50, 30, 100, 200]
box_xyxy = [box_xywh[0], box_xywh[1], box_xywh[0] + box_xywh[2], box_xywh[1] + box_xywh[3]]
print(box_xyxy)

A [0, 0, 150, 230]

B [50, 30, 100, 200]

C [50, 30, 150, 230]

D [50, 30, 50, 30]

💡 Hint

Add width to x and height to y to get bottom-right corner.

The bottom-right corner is calculated by adding width to x and height to y, so (50+100, 30+200) = (150, 230).

❓ Model Choice

advanced

Choosing bounding box format for object detection model

You are designing an object detection model that predicts bounding boxes. Which bounding box representation is most suitable for regression output in a neural network?

A Top-left and bottom-right corners (x1, y1, x2, y2)

B Center coordinates with width and height (cx, cy, w, h)

C Polygon points around the object

D Binary mask of the object

💡 Hint

Think about what is easier for a network to predict smoothly.

Center coordinates with width and height are continuous and easier for regression, making them suitable for neural network outputs.

❓ Metrics

advanced

Evaluating bounding box predictions with IoU

What does an Intersection over Union (IoU) score of 0.75 between a predicted bounding box and the ground truth indicate?

A The predicted box overlaps 75% with the ground truth box

B The predicted box is 75% larger than the ground truth box

C The predicted box is 75% smaller than the ground truth box

D The predicted box and ground truth box do not overlap

💡 Hint

IoU measures overlap area divided by union area.

IoU is the ratio of the overlapping area to the combined area of both boxes. A score of 0.75 means 75% overlap.

🔧 Debug

expert

Debugging bounding box normalization error

Given an image of size 200x100 pixels and a bounding box [50, 20, 150, 80] in XYXY format, a developer tries to normalize the box coordinates by dividing by image width and height as follows: normalized_box = [50/100, 20/200, 150/100, 80/200] What is the main issue with this normalization approach?

A The box coordinates are not in XYXY format

B The box coordinates exceed image dimensions

C The width and height are swapped in division

D The normalization should use width for x-coordinates and height for y-coordinates

💡 Hint

Check which coordinates correspond to width and height axes.

The code incorrectly divides x-coordinates by image height (100) and y-coordinates by image width (200). X-coordinates should be divided by image width (200), and y-coordinates by image height (100).