Maswali yanayounganishwa na mwongozo · Ngumu Kiwango

Maswali ya Kugundua Kibadilishaji cha DETR

Jaribio la hoja za kifaa cha DETR, ulinganishaji wa Kihungari, shabaha zisizo na kitu, uainishaji na hasara za kisanduku, na sababu ya kigunduzi chake chenye msingi kuepuka NMS.

Njia za mwongozo zinazohusianaUtambuzi wa Kibadilishaji cha Detr
Swali 1 ya 8

Je, DETR huondoa nini ikilinganishwa na vigunduzi vya kitamaduni kama vile Faster R-CNN?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Je, DETR huondoa nini ikilinganishwa na vigunduzi vya kitamaduni kama vile Faster R-CNN?
    • Vipengee vilivyoundwa kwa mkono kama vile masanduku ya nanga na ukandamizaji usio wa kiwango cha juu
    • Uti wa mgongo wa CNN kabisa
    • Haja ya data yoyote ya mafunzo
    • Sanduku za kufunga kama umbizo la towe
  2. DETR hutumia algorithm gani kulinganisha utabiri na vitu vya ukweli wakati wa mafunzo?
    • k-inamaanisha kuunganisha
    • Kuongeza gradient
    • Algorithm inayolingana ya Hungarian (bipartite).
    • Utafutaji wa boriti
  3. 'maswali ya kitu' ya DETR ni nini?
    • Vidokezo vya maandishi vilivyochapwa na mtumiaji
    • Sanduku za nanga za saizi zisizohamishika
    • Sampuli za saizi kutoka kwa picha
    • Seti isiyobadilika ya upachikaji zilizofunzwa, avkodare hubadilika kuwa utambuzi
  4. DETR inachanganya usanifu gani wa jumla?
    • Mtandao unaotumika mara kwa mara pamoja na mti wa maamuzi
    • Uti wa mgongo wa CNN pamoja na kisimbaji cha kusimba cha transfoma
    • GAN safi
    • Kiainishi cha k-karibu-jirani
  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. Je, ni udhaifu gani unaojulikana wa DETR asili?
    • Muunganiko wa polepole wa mafunzo na utambuzi dhaifu wa vitu vidogo
    • Haikuweza kukimbia kwenye GPU
    • Haihitaji data iliyo na lebo
    • Ilifanya kazi tu kwenye picha za kijivu
  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