Kediska Kala-duwanaanshaha Dila iyo Arous

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Muxuu heerka fidinta ka weyn 1 ku sameeyaa kernel convolution?
   • Waxa ay gelisaa daldaloolo u dhexeeya miisaanka kernel-ka si ay u daboolaan aag ballaaran
   • Waxay ku darsataa miisaan badan oo la baran karo kernel-ka
   • Waxa ay kernel-ku hoos u dhigtaa hal dhibic
   • Waxay u beddeshaa kernel-ka hawl-wadaag
2. 3x3 kernel oo leh xadiga fidinta ee 2 ayaa si wax ku ool ah u cabbiraya meeqa gobol?
   • 3x3
   • 5x5
   • 7x7
   • 9x9
3. Waa maxay sababta isfaham la ballaadhinayo ay si gaar ah faa'iido ugu leeyihiin qaybinta semantiga?
   • Waxay u yareeyaan sawirka hal calaamad
   • Waxay meesha ka saarayaan baahida loo qabo xog kasta oo tababar ah
   • Waxay balaadhiyaan goobta soo dhaweynta iyagoo sare u qaadaya xallinta khariidadda astaanta
   • Waxay kaliya ku shaqeeyaan sawirada cawl
4. Muxuu ka yimid ereyga atrous?
   • Eray Laatiinka ah oo lakabka ah
   • Soo gaabinta habka tababarka
   • Magaca hindiseheeda
   • Faransiisku waa surwaal, oo macnaheedu yahay godadka
5. WaveNet dhexdeeda, maxay isu-garabyada kala-baxa ah ugu qiimo badan yihiin?
   • Waxay abuuraan sawirro halkii ay ka ahaan lahaayeen maqal
   • Waxay si hufan u qabtaan ku-tiirsanaanta fogaanta fog ee qaabka mowjadda maqalka si hufan
   • Waxay ka saareen dhammaan aamusnaanta duubista
   • Waxay maqalka u beddelaan qoraal toos ah
6. Dhibaato intee le'eg ayaa iman karta haddii aad isku dhejiso iskudhacyada fidsan oo leh qiime keliya oo liidata la doortay?
   • Shabakadu waxa ay noqotaa mid aad u yar oo lagu tababari karo
   • Sawirku si joogto ah ayaa loo rogaa
   • Farshaxannada isku dhejinta, halkaasoo boosaska wax gelinta qaarkood aan weligood la muunad qaadin
   • Dhammaan miisaanku waxay noqonayaan eber
7. How does Atrous Spatial Pyramid Pooling combine multiple scales in DeepLab?
   • It runs several dilation rates in parallel and fuses their outputs
   • It chooses one fixed dilation rate and discards the rest
   • It converts the image into a single global class before convolution
   • It replaces every convolution with a larger dense kernel
8. For one input channel and one output channel, how many kernel weights does a dilated 3×3 filter contain, excluding a bias?
   • 25, because dilation 2 spans a 5×5 region
   • 5, one for each position along the expanded width
   • 18, because dilation 2 doubles the number of weights
   • 9, because dilation changes spacing rather than the 3×3 weight count