Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

StyleGAN Architecture adanwo

Review StyleGAN mapping and synthesis, layerwise style mixing, stochastic noise, and the architectural changes in StyleGAN2 and StyleGAN3.

Jẹmọ itọsọna awọn ọnaStylegan faaji
Ibeere 1 ti 8

Kini idi ti nẹtiwọọki aworan agbaye ti StyleGAN?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. Kini idi ti nẹtiwọọki aworan agbaye ti StyleGAN?
    • Lati ṣe lẹtọ boya aworan kan jẹ gidi tabi iro
    • Lati ṣe agbega awọn aworan si ipinnu giga
    • Lati compress ik aworan fun ibi ipamọ
    • Lati yi iwifun z pada si aaye agbedemeji ti o yapa diẹ sii W
  2. Ninu StyleGAN atilẹba, bawo ni a ṣe itasi ara sinu awọn maapu ẹya?
    • Nipa sisopọ fekito wiwaba si ẹbun kọọkan
    • Nipasẹ Iṣatunṣe Apeere Adaṣe (AdaIN)
    • Nipasẹ a ik softmax Layer
    • Lilo max pooling
  3. Kini nẹtiwọọki iṣelọpọ bẹrẹ lati dipo fekito wiwaba?
    • A kẹkọọ ibakan tensor
    • Aworan oju gidi kan
    • Ariwo funfun funfun ni iwọn kanna bi abajade
    • Maapu eti kan
  4. Kini awọn aṣa ti n ṣatunṣe ni awọn ipele isokuso (ipinnu-kekere) ni akọkọ iṣakoso?
    • Fine ara sojurigindin ati freckles
    • Awọ ti awọn piksẹli kọọkan nikan
    • Awọn abuda ipele giga bi iduro ati apẹrẹ oju
    • Aworan faili kika
  5. Iṣoro wo ni StyleGAN2 ṣe atunṣe ojulumo si atilẹba?
    • O ṣafikun ifọrọranṣẹ-si-aworan
    • O yọkuro awọn ohun-ọṣọ bii blob abuda ti o fa nipasẹ AdaIN
    • O pa iwulo fun eleyatọ kuro
    • O ṣe ikẹkọ ni kikun ti ko ni abojuto fun igba akọkọ
  6. Kini 'dapọ ara' ni StyleGAN?
    • Dapọ awọn datasets meji ṣaaju ikẹkọ
    • Iwọn aropin awọn abajade eleyatọ
    • Dapọ ohun pẹlu aworan ti ipilẹṣẹ
    • Lilo oriṣiriṣi w vectors ni awọn ipele oriṣiriṣi lati darapo awọn abuda lati awọn aworan meji
  7. What role do separate per-layer noise inputs play in original StyleGAN synthesis?
    • They replace the mapping network's entire style representation
    • They determine the file's compression format
    • They supply stochastic fine detail, such as varying hair strands or freckles
    • They provide the discriminator's real-image labels
  8. What visual behavior did StyleGAN3's alias-free design target?
    • Texture sticking and aliasing that make details move unnaturally as image content changes
    • The absence of a text tokenizer in original StyleGAN
    • A requirement that every output be a grayscale photograph
    • The need to store generated images as audio waveforms