Workshop (P2P)

What does YOLO stand for?

• You Only Like Optimization

• You Only Look Once

• You Only Love Once

• Your Objective Label Output

What is the main task of YOLO in tracking pipeline?

• Classify players into teams

• Detect players positions

• Detect objects (players) in each frame

• Predict players future position

Which output does YOLO NOT produce directly?

• Confidence score

• Class label

• Bounding boxes coordinates

• Object ID for tracking

What is BYTETracker role in our pipeline?

• Detect object (players)

• Associate detections across frames

• Resize images

• Detect objects

What is the correct order in our pipeline?

• 1. 1. Detect and track players simultaneously

  2. 2. Classify players into teams

• 1. 1. Players detection

  2. 2. Classify players into teams

  3. 3. Track detected players

• 1. 1. Detect players

  2. 2. Track players

  3. 3. Team classification

• 1. 1. Classify players into teams based on their jersey color

  2. 2. Track those players across frames

What happens if YOLO fails to detect a player in a frame?

• BYTETracker may lose track of that player

• The player is manually added

• The frame is skipped

• The player position in the following frames will be decided by Kalman Filter

Which parameter controls how many times the model will see the entire training set?

• Learning Rate

• Batch Size

• Epochs

• IoU Threshold

What format must the detections be in for BYTETracker to use them?

• [frame, x1, x2, y1, y2]

• [frame, x1, x2, y1, y2, score, class]

• [frame, x, y, w, h, score, class]

• [frame, x1, y1, x2, y2, score, class]

What is the maximum number of classes can you use in a YOLO training dataset?

• 3

• 7

• There is no limit

• 4

Messi or Ronaldo? 🐐

• Ronaldo

• Both are great players, we should enjoy as long as they are still playing! ❤

• Messi

Answer explanation

Suiiiiiiiiiiii. We all know he is the GOAT