Kedis ku xidhan hagaha · Dhexdhexaad Heerka

Kediska Soo Celinta Transformer SwinIR

Baadh sida SwinIR ay ula qabsato blocks-ka beddelka daaqadaha ee xallinta sare, diidmada, iyo ka saarista qalab-samaynta.

Wadooyinka haga ee la xidhiidhaSoo Celinta Transformer Swinir
Su'aal 1 ee 8

Waa maxay qaab-dhismeedka beddelka ee SwinIR ku salaysan yahay?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Waa maxay qaab-dhismeedka beddelka ee SwinIR ku salaysan yahay?
    • Transformer-ka Swin
    • Beddelka Aragga (ViT) kaliya
    • BERT
    • GPT
  2. Waa maxay habka dareenka xudunta u ah SwinIR?
    • Feejignaanta dhammaan-pixel-ka caalamiga ah
    • Is-fiirsi ku salaysan daaqada oo isbedelay
    • Fiiro gaar ah sawirka
    • Fiiro gaar ah ma leh, kaliya qalalaasaha
  3. Maxaa loo yaqaan SwinIR's blocks-spiritual?
    • Is-hor-istaagga Cufan
    • Baloogyada Haraaga ah ee leexday
    • Baloogyada beddelka dabaasha ee hadhaaga ah (RSTB)
    • Kaabsal Blocks
  4. Waa maxay sababta dareenka ku salaysan daaqada uu uga waxtar badan yahay dareenka caalamiga ah halkan?
    • Waxay isticmaashaa kanaalo midab yar
    • Waxay ka booddaa tallaabada dib-u-dhiska
    • Waxay ka shaqeysaa cidhifyada
    • Qiimaheedu waxa uu u koraa si toos ah oo leh bedka sawirka halkii uu ka ahaan lahaa afar gees
  5. What are the three main stages in SwinIR’s restoration architecture?
    • Shallow feature extraction, deep feature extraction, and image reconstruction
    • Text tokenization, translation, and speech synthesis
    • Noise injection, denoising, and audio decoding
    • Camera sensing, object tracking, and route planning
  6. Waa maxay faa'iidada waxtarka leh ee SwinIR ka muujisay tartamayaasha CNN?
    • Uma baahnayn xog tababar
    • Waxay gaadhay natiijooyin casri ah oo leh cabbirro aad u yar
    • Waxay ku socotay iyada oo aan wax qalalaase ahi dhicin
    • Waxay meesha ka saartay baahida GPU-yada
  7. What components does a SwinIR Residual Swin Transformer Block combine?
    • A text tokenizer and a language-model decoder
    • A GAN discriminator and a vocoder
    • Swin Transformer layers with residual connections and a convolution
    • A camera sensor and a LiDAR encoder
  8. Which image-restoration task besides super-resolution did the SwinIR paper evaluate?
    • Aqoonsiga hadalka
    • Turjumaada mashiinka
    • Protein sequence design
    • JPEG compression-artifact reduction