Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Ojú Sisan adanwo

Test optical-flow vectors, brightness constancy, RAFT, occlusion, motion limits, and task-specific validation. Check evidence and limits.

Jẹmọ itọsọna awọn ọnaSisan Opitika
Ibeere 1 ti 8

Kí ni opitika sisan kosi ti siro?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. Kí ni opitika sisan kosi ti siro?
    • Iṣipopada ti awọn piksẹli laarin awọn fireemu itẹlera
    • Awọn awọ otutu ti kọọkan fireemu
    • Nọmba awọn nkan ni aworan kan
    • Fọọmu igbi ohun ti o tẹle fidio kan
  2. Kini arosinu 'imọlẹ nigbagbogbo' ti a lo ninu ṣiṣan opiti Ayebaye?
    • Gbogbo fireemu gbọdọ ni imọlẹ apapọ kanna
    • Aaye gbigbe kan tọju aijọju imọlẹ kanna kọja awọn fireemu
    • Imọlẹ gbọdọ pọ sii ni imurasilẹ lori fidio kan
    • Fidio dudu-funfun nikan ni a le ṣe itupalẹ
  3. Which method uses a 4D correlation volume and recurrent updates for optical flow?
    • Tesseract
    • Mask R-CNN
    • RAFT
    • Swin patch merging
  4. Why can classic optical-flow assumptions fail in fast motion?
    • Because frames contain no pixels
    • Because flow is a text-only task
    • Because all motion is uniform
    • Large displacement, occlusion, or brightness changes violate local smoothness or brightness-constancy assumptions
  5. Ohun elo wo ni o gbẹkẹle iṣiro iṣiro laarin awọn fireemu?
    • Fidio fireemu interpolation fun dan o lọra išipopada
    • Akọtọ-ṣayẹwo iwe ọrọ kan
    • Tito awọn fọto nipasẹ iwọn faili
    • Yiyipada ọrọ si ọrọ
  6. Kini iṣoro 'occlusion' ni ṣiṣan opitika?
    • Nigbati fidio ba tan imọlẹ pupọ
    • Nigbati piksẹli ba farapamọ tabi fi han, nitorina ko si baramu to wulo laarin awọn fireemu
    • Nigbati awọn fireemu oṣuwọn jẹ gangan 30fps
    • Nigbati awọn fidio meji ba dapọ
  7. What is the output of dense optical flow?
    • One class label for the whole clip
    • A 3D mesh with metric scale
    • A displacement vector estimate for each image pixel
    • A transcript of the audio
  8. What is one reason flow can be uncertain in a textureless region?
    • Every pixel is assigned a unique object name
    • The region has too many labels
    • Optical flow only works on text
    • There may be too little local visual information to identify a unique match