Kvíz spojený s průvodcem · Těžko úroveň

Animace portrétu z kvízu o jedné fotografii

Testuje přenos pohybu založený na klíčových bodech, manipulaci s okluzí a možnosti designu SadTalker a LivePortrait pro animaci obličeje z jedné fotografie.

Související vodicí cestyPortrét Animace Z Jedné Fotografie
Otázka 1 z 8

How does the First Order Motion Model obtain its keypoints?

Náhled každé otázky8 otázky

Odpovědi a vysvětlení se zobrazují, jak budete v kvízu.

  1. How does the First Order Motion Model obtain its keypoints?
    • From manually labeled facial landmarks
    • From a 3D scan of the person
    • It learns them without labels during training
    • From the audio track
  2. What is the purpose of the occlusion map?
    • To mark regions the source image cannot supply, so they must be painted in
    • To blur the background on purpose
    • To detect the speaker's language
    • To choose the output frame rate
  3. What is the benefit of relative motion transfer?
    • It doubles the frame rate
    • The source keeps its own face shape instead of taking on the driver's
    • It removes the need for a driving video
    • It makes the head turn farther
  4. What does SadTalker predict from the audio?
    • Raw pixel values of the mouth
    • A new background
    • Textový přepis
    • 3D Morphable Model coefficients for expression and head pose
  5. What does LivePortrait's stitching module do?
    • It pastes the animated face back into the full image without misaligned shoulders
    • It merges two different people's faces
    • It stitches audio clips together
    • It converts video to GIF
  6. Why do large head turns usually fail in single-photo animation?
    • The GPU runs out of memory
    • Head turns are disabled by default
    • The photo has no information about unseen regions like the side of the head
    • Audio cannot describe rotation
  7. Why do keypoint methods warp feature maps rather than raw pixels?
    • Pixels cannot be moved
    • It lets the decoder repair stretching and fill unseen regions
    • Feature maps are always higher resolution
    • It removes the need for keypoints
  8. In implicit 3D keypoint methods, what does the driving frame contribute?
    • The canonical keypoints encoding identity
    • The background texture
    • The skin color
    • Head rotation, translation and expression deformations