quiz bu lëkkale ak tegtal · Dëgër Niveau
Laajteb konvolusioŋ yuñ mëna soppi
Saytu liñu jàng ci jël misaal yi, interpolaasioŋ bilineer, modulasioŋ, ak boole gox yi ci reso konvolusioŋ yuñ mëna soppiku ak gis.
Laaj 1 de 8
Lan la convolusioŋ buñ mëna deforme di yokk suñu ko méngale ak convolusioŋ buñ miin?
Xoolaatal laaj bu nekk8 laaj
Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.
- Lan la convolusioŋ buñ mëna deforme di yokk suñu ko méngale ak convolusioŋ buñ miin?
- Dayo kernel bu gëna mag
- Jàng nañu ay offset 2D yuy soppi barabu jël misaal bu nekk
- ab pooling jéego ginaaw bépp couche
- Dropout aleatoire ci poid yi
- naka lañuy defaree ay kompensaasioŋ ngir jël misaal ci benn convolusioŋ buñ mëna deforme?
- Ingénieur bi moo leen di kode
- Dañuy nekk ay constant yu bawoo ci tablo seetlu
- Ñu ngi leen kopie ci kaadar bi njëkk
- Beneen couche de convolution moo leen di wax luy waaja am ci liñu dugal
- Ndax offset yi ñuy séentu dañuy faral di nekk fractionnel, naka lañuy jëlee valeur pixel yi ci position yooyu?
- interpolaasioŋ bilineer
- Dëkkandoo bi gëna jege
- pooling bu gëna mag
- Dañu leen sànni
- Lan la ConvNets v2 (2018) yokk ci li njëkk?
- Lëkkaloo yuy dellusi
- ab fonction perte bu bees
- Poids modulation ci misaal bu nekk dafay xayma li poñ bu nekk mëna def
- Dindi bépp offset
- Lan moo waral convolution yuñ mëna soppi dañu am njariñ ci mbir yu am jëmm ju jaxasoo?
- Seeni champ réceptif mën na deforme ngir topp contour objet bi
- Dañuy yokk dayo nataal bi
- Dañuy kompresse dayo model bi
- Dañuy soppi nataali melo yi ci grayscale
- Ban xeetu gis-gis bu bees mooy jëfandikoo xalaat deformable ci bàyyi xel ci transformatër?
- Resnet-50
- AlexNet
- U-Net
- DETR buñu mëna soppiku
- For an N-point two-dimensional deformable kernel, how many offset components are predicted at each output location?
- N: one class probability per point
- N squared: one displacement for every pair of points
- 2N: one horizontal and one vertical displacement per point
- Two total, shared by all N points
- Where does deformable RoI pooling apply adaptive offsets?
- Only to the detector's output class names
- To sampling locations within a proposed object's region features
- To the ground-truth box coordinates before annotation
- Only to the image file's compression blocks