Chapter 27: Computer Vision and Video Analysis - Quiz

Instructions

This quiz contains 25 questions covering the key concepts from Chapter 27. Select the best answer for multiple-choice questions. Short answer questions should be answered in 2-4 sentences.

Multiple Choice Questions

Question 1

The standard frame rate for NBA tracking systems is approximately: - A) 10 fps - B) 25 fps - C) 60 fps - D) 120 fps

Question 2

Which tracking technology is used in official NBA games? - A) GPS tracking - B) RFID tags - C) Optical camera tracking - D) Accelerometers

Question 3

Pose estimation systems typically identify how many keypoints on the human body? - A) 5-8 - B) 10-12 - C) 17-25 - D) 50+

Question 4

The main challenge of tracking the basketball compared to players is: - A) The ball is too small - B) The ball moves faster and is frequently occluded - C) The ball has no texture - D) Cameras cannot see the ball

Question 5

YOLO (You Only Look Once) is primarily used for: - A) Pose estimation - B) Object detection - C) Action recognition - D) Depth estimation

Question 6

Optical flow in video analysis measures: - A) Player heights - B) Motion between consecutive frames - C) Court boundaries - D) Ball spin

Question 7

The primary purpose of homography in basketball video analysis is: - A) Color correction - B) Mapping camera view to court coordinates - C) Player identification - D) Ball tracking

Question 8

For real-time applications, which pose estimation system is typically preferred? - A) OpenPose - B) AlphaPose - C) MediaPipe - D) DensePose

Question 9

Second Spectrum became the official tracking provider for the NBA in: - A) 2010 - B) 2013 - C) 2017 - D) 2021

Question 10

When a player is occluded by another player, tracking systems typically: - A) Mark the position as unknown - B) Interpolate the position - C) Stop tracking that player - D) Use GPS backup

Question 11

Convolutional Neural Networks (CNNs) are primarily used for: - A) Sequence modeling - B) Extracting spatial features from images - C) Tracking player positions - D) Calculating statistics

Question 12

LSTMs are particularly useful in basketball video analysis for: - A) Image classification - B) Temporal sequence modeling - C) Object detection - D) Camera calibration

Question 13

The typical spatial accuracy of optical tracking systems is approximately: - A) Within a few inches - B) Within a few feet - C) Within a few yards - D) Varies randomly

Question 14

Jersey number recognition is challenging because: - A) Players don't have numbers - B) Numbers are small, move quickly, and get occluded - C) All jerseys look the same - D) Cameras are too far away

Question 15

Multi-camera systems improve tracking accuracy primarily through: - A) Higher frame rates - B) Better resolution - C) Reducing occlusion and improving triangulation - D) Color accuracy

Question 16

Graph Neural Networks (GNNs) are increasingly used in basketball for: - A) Image enhancement - B) Modeling player interactions and formations - C) Ball tracking - D) Jersey recognition

Question 17

The term "Part Affinity Fields" is associated with: - A) Ball tracking - B) OpenPose's method for connecting body parts - C) Camera calibration - D) Action recognition

Question 18

Which metric is commonly used to evaluate object detection models? - A) Accuracy - B) mAP (mean Average Precision) - C) F1 score - D) AUC-ROC

Question 19

Transfer learning in basketball video analysis typically involves: - A) Using models trained on general video datasets - B) Transferring players between teams - C) Moving data between servers - D) Converting video formats

Question 20

The main advantage of edge computing for basketball analysis is: - A) Lower cost - B) Reduced latency for real-time applications - C) Better accuracy - D) Simpler deployment

Short Answer Questions

Question 21

Explain why computer vision for basketball is more challenging than for sports like soccer or tennis. Mention at least three specific challenges.

Your Answer:

Question 22

Describe how you would use pose estimation to analyze and improve a player's free throw shooting form. What specific keypoints and angles would you measure?

Your Answer:

Question 23

A coach wants to know how well the team spaces the floor on offense. Describe how you would use tracking data to create a meaningful "spacing quality" metric.

Your Answer:

Question 24

Explain the trade-off between accuracy and latency in real-time basketball video analysis. How would requirements differ for in-game coaching versus post-game analysis?

Your Answer:

Question 25

Discuss the privacy and ethical considerations of using computer vision systems that can identify and track individual players continuously throughout a basketball game.

Your Answer:

Answer Key

Multiple Choice Answers

  1. B - NBA tracking systems operate at 25 fps, capturing player and ball positions 25 times per second.

  2. C - The NBA uses optical camera tracking (currently Second Spectrum) rather than wearable devices.

  3. C - Most pose estimation systems identify 17-25 keypoints (COCO has 17, BODY_25 has 25).

  4. B - The ball moves at high speeds (50+ mph on passes) and is frequently hidden by players' bodies.

  5. B - YOLO is an object detection algorithm known for real-time performance.

  6. B - Optical flow measures the apparent motion of pixels between consecutive frames.

  7. B - Homography transforms camera perspective to a top-down court view.

  8. C - MediaPipe is optimized for real-time applications on various devices.

  9. C - Second Spectrum became the official NBA tracking provider in 2017.

  10. B - Tracking systems typically interpolate positions during occlusion using motion models.

  11. B - CNNs are designed to extract spatial features from image data.

  12. B - LSTMs (Long Short-Term Memory networks) model sequential/temporal data.

  13. A - Modern optical tracking achieves accuracy within a few inches.

  14. B - Jersey numbers are small relative to image size, move rapidly, and are often partially visible.

  15. C - Multiple cameras reduce occlusion issues and enable 3D triangulation.

  16. B - GNNs model relationships between entities, ideal for player interactions.

  17. B - PAFs are OpenPose's method for associating detected body parts with individuals.

  18. B - mAP (mean Average Precision) is the standard metric for object detection evaluation.

  19. A - Transfer learning uses models pre-trained on ImageNet or Kinetics datasets.

  20. B - Edge computing processes data locally, reducing latency for real-time applications.

Short Answer Rubric

Question 21 - Should mention: (1) fast player movement and rapid direction changes, (2) frequent occlusion in tight spaces, (3) similar uniforms on same team, (4) ball speed and frequent hand changes, (5) indoor lighting variations across arenas.

Question 22 - Should discuss: elbow angle at release, knee bend, follow-through wrist position, shoulder alignment, balance/center of mass. Measure angles between keypoints (shoulder-elbow-wrist). Compare to established "good form" benchmarks.

Question 23 - Should describe: calculate distances between all five players, potentially use convex hull area, measure distance from optimal positions, consider dynamic spacing during play vs static. Weight by offensive context.

Question 24 - Should explain: real-time requires low latency (<100ms), may sacrifice accuracy. In-game coaching needs immediate feedback (simple metrics). Post-game analysis can use complex processing (detailed breakdowns). Model complexity and hardware determine tradeoff.

Question 25 - Should address: player consent for tracking, data ownership and access control, potential for surveillance beyond sport, biometric data sensitivity, ensuring data is used only for stated purposes, regulatory compliance.