Ikibazo gihuza ibibazo · Hagati Urwego
Ikibazo cya Autoencoders Ikibazo
Ongera usubiremo MAE ya masike yo hejuru yibikorwa byo kongera kubaka, igishushanyo mbonera cya asimmetrike-decoder, hamwe nimpapuro zasuzumwe mbere yo gushiraho.
Ikibazo 1 Bya 8
Ni ubuhe butumwa bwibanze bwo kwiyobora muri Masked Autoencoder?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Ni ubuhe butumwa bwibanze bwo kwiyobora muri Masked Autoencoder?
- Kongera kubaka pigiseli yama patch yari yihishe
- Gutondekanya buri shusho mubyiciro byanditse
- Guteganya ibizakurikiraho muburyo bwa videwo
- Guhuza ishusho ninyandiko yanditse
- Hafi yikihe gice cyibice byamashusho MAE yumwimerere isanzwe yihishe?
- Abagera kuri 15%
- Abagera kuri 75%
- Abagera kuri 35%
- Abagera kuri 95%
- Kuki kodegisi ya MAE itunganya gusa ibiboneka (bidafite ibimenyetso)?
- Kuberako ibipfundikizo bipfutse ubusa nta bara bifite akamaro
- Kurinda icyitegererezo kutabona amashusho nyayo
- Kugira ngo imyitozo yihute kandi yibuke neza
- Kuberako decoder idashobora gusoma kodegisi ya posisiyo
- Bigenda bite kuri decoder nyuma ya MAE kwitegura birangiye?
- Ihinduka ibyiciro nyamukuru
- Irahagaritswe kandi ikoreshwa mugukora amashusho mashya
- Isimbuza kodegisi yose
- Yajugunywe; kodegisi ikoreshwa kumurimo wo hasi
- Nibihe bikorwa byigihombo MAE ikoresha muburyo bwo kwiyubaka?
- Hagati ya kwaduka ikosa kuri maskike-patch pigiseli
- Kwambukiranya-hejuru ya label y'ibyiciro
- Gutandukanya AmakuruNCE igihombo
- Igihombo cyo kuvangura
- Igishushanyo cya 'asimmetric' cya MAE bivuga iki?
- Ingano itandukanye yishusho yo kwinjiza no gusohoka
- Kodegisi iremereye ihujwe na decoder yoroheje
- Gukoresha imibare ibiri itandukanye mumahugurwa
- Guhisha igice cyibumoso cya buri shusho
- On which large image dataset did the original MAE paper pretrain its main vision models?
- MS COCO Captions only
- Ubuntu
- ImageNet-1K
- The Netflix Prize ratings set
- What does the MAE decoder receive for image patches hidden from the encoder?
- The original RGB pixels without masking
- Only class labels from ImageNet
- A text caption for the image
- Learned mask tokens with positional embeddings