Útmutatóhoz kapcsolódó kvíz · Közepes Szint

Hand Pose Estimation and Hand Tracking Quiz

Tests landmarks, video tracking, gesture interpretation, failure modes, and human-centered validation of hand pose systems and apps.

Kapcsolódó útmutató utakHand Pose Estimation And Hand Tracking
kérdés 1 -ból 8

What does MediaPipe Hand Landmarker output for each detected hand?

Tekintse meg az összes kérdést8 kérdéseket

A válaszok és magyarázatok megjelennek a kvíz kitöltésekor.

  1. What does MediaPipe Hand Landmarker output for each detected hand?
    • A guarantee of the person’s intended gesture
    • A hand image with all occlusion removed
    • A set of 21 hand landmarks, including key hand points
    • A complete sign-language translation for every frame
  2. How does hand-pose estimation differ from hand tracking?
    • Estimation finds keypoints in an image; tracking associates detections over time across frames
    • Estimation identifies user intent while tracking measures emotion
    • Tracking works only on still images
    • They are identical terms for the same one-frame result
  3. Before landmarks control a gesture interface, what must the application define?
    • A wrist point that infers every hand action
    • Patterns mapped to actions, with user feedback
    • Coordinates that automatically encode the user’s command
    • A universal gesture vocabulary supplied by the landmark model
  4. When two visible hands cross in a tracked sequence, which identity error can occur?
    • The hand keypoints can become noisier while the same track IDs remain attached
    • The tracker may temporarily lose one hand but keep its last identity until detection resumes
    • The gesture classifier may confuse a pose while the hand-to-track association remains stable
    • The tracker can swap which physical hand its two track IDs refer to
  5. Why test a live gesture pipeline at its intended frame rate?
    • Latency and frame sampling can affect tracking continuity and action timing
    • Camera frame rate sets temporal sampling, but cannot affect action timing once points are detected
    • Display refresh rate alone determines tracking latency regardless of camera input
    • The same smoothing window preserves timing at every capture frame rate
  6. What does a landmark model alone not provide?
    • A set of landmarks in image coordinates
    • Detection results for visible hand poses
    • A guaranteed interpretation of what a person intends
    • The estimated locations of hand keypoints
  7. Which evaluation is most useful for a hand-controlled app?
    • Check whether the camera lens is a wide-angle model
    • Measure localization error, misses, identity switches, latency, and user comfort
    • Count only the number of keypoints in the model
    • Measure training accuracy on one clean video
  8. For a high-impact gesture action, which safeguards should the app provide?
    • Smooth a pose across frames and show feedback, then trigger automatically without a cancel path
    • Require a one-time setup confirmation and trigger later actions from a single high-confidence frame
    • Provide pause and undo controls, but allow a single-frame gesture to trigger a consequential action
    • Require stable, user-confirmed input and provide pause or undo controls