Ikibazo gihuza ibibazo · Biragoye Urwego

Ikibazo cya DETR Guhindura Ikibazo

Ikizamini cya DETR cyibibazo, guhuza Hongiriya, nta ntego-intego, gutondekanya no gutakaza agasanduku, nimpamvu igikoresho cyacyo gishingiye kuri deteter yirinda NMS.

Inzira zijyanye no kuyoboraKumenya Guhindura
Ikibazo 1 Bya 8

Niki DETR ikuraho ugereranije na detector gakondo nka Byihuta R-CNN?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Niki DETR ikuraho ugereranije na detector gakondo nka Byihuta R-CNN?
    • Ibikoresho byakozwe n'intoki nkibisanduku bya ankeri hamwe no guhagarika ntarengwa
    • Umugongo wa CNN rwose
    • Gukenera amakuru ayo ari yo yose
    • Guhuza agasanduku nkibisohoka
  2. Niyihe algorithm DETR ikoresha kugirango ihuze ibyahanuwe nibintu-byukuri mugihe cy'amahugurwa?
    • k-bisobanura guhuriza hamwe
    • Kongera imbaraga
    • Hongiriya (bipartite) ihuye na algorithm
    • Gushakisha ibiti
  3. Nibihe 'DETR yibibazo'?
    • Inyandiko isaba yanditswe numukoresha
    • Agasanduku k'udusanduku twinshi
    • Pixels yatoranijwe kuva ishusho
    • Igice gihamye cyo kwiga gushiramo decoder ihinduka mubushakashatsi
  4. Ni ubuhe bwubatsi muri rusange DETR ihuza?
    • Umuyoboro usubiramo wongeyeho igiti cyemezo
    • Umugongo wa CNN wongeyeho transformateur encoder-decoder
    • GAN
    • A k-hafi-umuturanyi utondekanya
  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. Niyihe ntege izwi ya DETR yumwimerere?
    • Buhoro buhoro guhugura hamwe no gutandukanya ibintu bito
    • Ntishobora gukora kuri GPUs
    • Ntabwo byasabye ko habaho amakuru yanditse
    • Cyakoraga gusa kumashusho yimyenda
  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