Feature Kediska Shabakadda Ahraamta

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Mushkiladda xudunta u ah ee Shabakadda Ahraamta Muujinta ay ugu horrayn xalliso?
   • Ogaanshaha walxaha ka muuqda miisaan aad u kala duwan
   • Yaraynta tirada sawirada tababarka ee loo baahan yahay
   • Ka saarida sanduuqyada xidhitaanka ee nuqul ka mid ah ogaanshaha ka dib
   • Moodooyinka isku-buufinta ee geynta mobilada
2. Marka la barbar dhigo dhismaha Ahraamta sawirka (cabbirka iyo dib u socodsiinta shabakada), FPN inta badan faa'iido ayay leedahay sababtoo ah:
   • Uma baahna shabakad laf dhabar haba yaraatee
   • Dib u isticmaal sifada mar hore lagu xisaabiyay shabakada dhexdeeda, isaga oo kaydinaya xisaabinta
   • Waxay meesha ka saaraysaa baahida loo qabo qalalaase kasta
   • Kaliya waxay ku shaqeysaa sawirro cawl leh
3. FPN, waa maxay doorka dariiqa kor-hooska?
   • Waxay tuuraysaa sifooyinka xallinta hooseeya
   • Waxay u beddeshaa sawirada si ay u muujiso vectors si loo kala saaro kaliya
   • Waxay sidda qani ahaan macne ahaan, sifooyin qoto dheer oo dib loogu celinayo xalal sare
   • Waxay qabataa xakamayn aan ugu badnaan
4. Sidee isku xirka dambe ee caadiga ah loogu daray astaamaha kor-sare?
   • Kanaalka isku xidhan-xikmad leh ka dibna kala badh la jaray
   • Cunsur la dhuftey oo caqli leh kadib softmax
   • Laga jaray si loo xisaabiyo khariidad kala duwan
   • Cunsur-caqli leh oo lagu daray ka dib 1x1 convolution ayaa toosiya kanaalada
5. Waa maxay sababta 3x3 convolution loogu dabaqay khariidad kasta oo la isku daray ee FPN?
   • Si loo hagaajiyo khariidadda oo loo dhimo kala-soocidda kor-u-qaadista
   • Si loo yareeyo xallinta goobta kala badh
   • Si loo kala saaro sawirka qaybo
   • Si toos ah loogu beddelo astaamaha isku-duwayaasha xidhidhiyaha
6. Nashqadadee ka dib ayaa ku kordhisay FPN oo leh la baran karo, isku dhafan laba jiho oo miisaan leh?
   • AlexNet
   • BiFPN, oo loo isticmaalo EfficientDet
   • LeNet-5
   • Word2Vec
7. How is a top-down FPN feature map typically resized before lateral fusion?
   • Downsampled to a single pixel
   • Rotated 90 degrees
   • Upsampled by a factor of two using nearest-neighbor interpolation
   • Converted into a text sequence
8. What does an FPN preserve by merging shallow high-resolution features with deeper semantic features?
   • Only the original image colors
   • Only the deepest layer’s low-resolution grid
   • The exact training labels without feature extraction
   • Fine spatial detail together with stronger semantic information