Kedis ku xidhan hagaha · Adag Heerka

Kediska Ogaanshaha Transformer DETR

Tijaabi DETR weydiimaha shayga, isbarbardhiga Hangari, yoolal aan shayga lahayn, kala soocida iyo khasaaraha santuuqa, iyo sababta baaraha ku salaysan uu uga fogaado NMS.

Wadooyinka haga ee la xidhiidhaOgaanshaha Transformer-ka Detr
Su'aal 1 ee 8

Maxay DETR meesha ka saartaa marka la barbar dhigo baarayaasha dhaqameed sida Degdega R-CNN?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Maxay DETR meesha ka saartaa marka la barbar dhigo baarayaasha dhaqameed sida Degdega R-CNN?
    • Qaybaha gacanta lagu naqshadeeyay sida sanduuqyada barroosinka iyo xakamaynta aan ugu badnaan
    • Laf-dhabarta CNN oo dhan
    • Baahida loo qabo xog kasta oo tababar ah
    • Sanduuqyada xidhidhiyaha sida qaabka wax soo saarka
  2. Algorithmkee ayay DETR isticmaashaa si ay u waafaqaan saadaasha walxaha runta ah inta lagu jiro tababarka?
    • k-macneheedu waa kooxaysi
    • Kordhinta tartiib tartiib ah
    • Hungarian (bipartite) algorithm iswaafajinta
    • Baadhitaanka Beam
  3. Waa maxay DETR's 'question questiones'?
    • Soo jeedinta qoraalka ee uu qoray isticmaaluhu
    • Sanduuqyada barroosinka ee cabbirada go'an
    • Pixels laga soo qaaday sawirka
    • Qayb go'an oo ku-xidhnaanshaha la bartay ayaa decoderiyuhu isu beddelaa ogaanshaha
  4. Waa maxay dhismaha guud ee DETR isku daray?
    • Shabakad soo noqnoqota oo lagu daray geed go'aan
    • Laf-dhabarka CNN oo ay weheliso cod-dedeeye koronto-bixiye
    • GAN saafi ah
    • Kala soocida deriska-k-ugu dhow
  5. How does DETR's training objective support inference without a separate NMS stage?
    • They make every possible duplicate prediction mathematically impossible
    • The model can only output one object per image
    • One-to-one matching and set losses discourage multiple predictions for the same object
    • A hidden NMS classifier supplies every training target
  6. Kee bay ahayd daciifnimada la yaqaan ee DETR asalka ah?
    • Isku xidhka tababarka oo gaabis ah iyo ogaanshaha shay yar oo daciif ah
    • Waxay ku socon kari wayday GPU-yada
    • Uma baahnayn xog sumadeysan
    • Waxay kaliya ka shaqeysay sawirro cawl ah
  7. What training label does original DETR assign to an object query that is not matched to a ground-truth object?
    • The largest visible object's class regardless of matching
    • The no-object class
    • A mandatory new ground-truth object invented for the query
    • The image's filename as a class label
  8. Which prediction errors does DETR's set loss penalize on matched object pairs?
    • Only the number of letters in the class name
    • Only the image's compression size
    • Only whether the training batch uses a GPU
    • Both the predicted class and bounding-box errors