본문

서브메뉴

Exploring Knowledge in Generative Models
Exploring Knowledge in Generative Models
Exploring Knowledge in Generative Models

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20260202105704
ISBN  
9798263308124
DDC  
004
저자명  
Bhattad, Anand.
서명/저자  
Exploring Knowledge in Generative Models
발행사항  
[Sl] : University of Illinois at Urbana-Champaign, 2024
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2024
형태사항  
93 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
주기사항  
Advisor: Forsyth, David A.
학위논문주기  
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2024.
초록/해제  
요약Generative models, such as StyleGAN, have demonstrated remarkable ability in producing realistic and controllable images. However, the underlying representations and mechanisms employed by these models remain largely unexplored. This thesis delves into the intrinsic properties and manipulability of StyleGAN, focusing on image relighting and decomposition. We begin by exploring the impact of image decompositions on image-based relighting. By analyzing the role of intrinsic image components such as reflectance, shading, and normals, we gain insights into the fundamental properties that contribute to realistic relighting. This understanding lays the foundation for our subsequent investigations into StyleGAN. Building upon these insights, we introduce StyLitGAN, a method that enables StyleGAN to generate scenes with novel lighting conditions. StyLitGAN produces realistic lighting effects, including cast shadows, soft shadows, inter-reflections, and glossy effects, without requiring labeled, paired, or CGI data. Moreover, it seamlessly extends to manipulating surface properties like colors and materials. Next, we present Make It So, a near-perfect GAN inversion technique that significantly outperforms previous state-of-the-art methods. Make It So can invert and relight real scenes, including out-of-domain images, demonstrating its generalizability and robustness. Finally, we uncover hidden gems within StyleGAN, providing strong evidence that it encodes easily accessible and accurate internal representations of familiar scene properties, known as ``intrinsic images," as defined by Barrow and Tenenbaum in their seminal work from 1978. We demonstrate that StyleGAN has encodings for intrinsic images such as reflectance, shading, and normals, which can be extracted and manipulated for various applications. Through our discoveries, we shed light on the implicit understanding of worldly knowledge present within generative models like StyleGAN. Our findings pave the way for improved manipulability, understanding, and refinement of generative models, with potential applications in computer vision, computational photography, computer graphics, and machine learning. This thesis contributes to the broader goal of leveraging generative models for advanced image manipulation and scene understanding tasks.
일반주제명  
Computer science
일반주제명  
Computer engineering
일반주제명  
Systems science
키워드  
Visual knowledge
키워드  
Generative models
키워드  
Intrinsic images
키워드  
Image decomposition
키워드  
Depth
키워드  
Albedo
키워드  
Segmentation
기타저자  
University of Illinois at Urbana-Champaign Computer Science
기본자료저록  
Dissertations Abstracts International. 87-05B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2024        us                              c    eng  d
■001000017361086
■00520260202105704
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798263308124
■035    ▼a(MiAaPQ)AAI32409900
■035    ▼a(MiAaPQ)124347
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a004
■1001  ▼aBhattad,  Anand.
■24510▼aExploring  Knowledge  in  Generative  Models
■260    ▼a[Sl]▼bUniversity  of  Illinois  at  Urbana-Champaign▼c2024
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2024
■300    ▼a93  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-05,  Section:  B.
■500    ▼aAdvisor:  Forsyth,  David  A.
■5021  ▼aThesis  (Ph.D.)--University  of  Illinois  at  Urbana-Champaign,  2024.
■520    ▼aGenerative  models,  such  as  StyleGAN,  have  demonstrated  remarkable  ability  in  producing  realistic  and  controllable  images.  However,  the  underlying  representations  and  mechanisms  employed  by  these  models  remain  largely  unexplored.  This  thesis  delves  into  the  intrinsic  properties  and  manipulability  of  StyleGAN,  focusing  on  image  relighting  and  decomposition.                        We  begin  by  exploring  the  impact  of  image  decompositions  on  image-based  relighting.  By  analyzing  the  role  of  intrinsic  image  components  such  as  reflectance,  shading,  and  normals,  we  gain  insights  into  the  fundamental  properties  that  contribute  to  realistic  relighting.  This  understanding  lays  the  foundation  for  our  subsequent  investigations  into  StyleGAN.                        Building  upon  these  insights,  we  introduce  StyLitGAN,  a  method  that  enables  StyleGAN  to  generate  scenes  with  novel  lighting  conditions.  StyLitGAN  produces  realistic  lighting  effects,  including  cast  shadows,  soft  shadows,  inter-reflections,  and  glossy  effects,  without  requiring  labeled,  paired,  or  CGI  data.  Moreover,  it  seamlessly  extends  to  manipulating  surface  properties  like  colors  and  materials.                        Next,  we  present  Make  It  So,  a  near-perfect  GAN  inversion  technique  that  significantly  outperforms  previous  state-of-the-art  methods.  Make  It  So  can  invert  and  relight  real  scenes,  including  out-of-domain  images,  demonstrating  its  generalizability  and  robustness.                        Finally,  we  uncover  hidden  gems  within  StyleGAN,  providing  strong  evidence  that  it  encodes  easily  accessible  and  accurate  internal  representations  of  familiar  scene  properties,  known  as  ``intrinsic  images,"  as  defined  by  Barrow  and  Tenenbaum  in  their  seminal  work  from  1978.  We  demonstrate  that  StyleGAN  has  encodings  for  intrinsic  images  such  as  reflectance,  shading,  and  normals,  which  can  be  extracted  and  manipulated  for  various  applications.                        Through  our  discoveries,  we  shed  light  on  the  implicit  understanding  of  worldly  knowledge  present  within  generative  models  like  StyleGAN.  Our  findings  pave  the  way  for  improved  manipulability,  understanding,  and  refinement  of  generative  models,  with  potential  applications  in  computer  vision,  computational  photography,  computer  graphics,  and  machine  learning.  This  thesis  contributes  to  the  broader  goal  of  leveraging  generative  models  for  advanced  image  manipulation  and  scene  understanding  tasks.
■590    ▼aSchool  code:  0090.
■650  4▼aComputer  science
■650  4▼aComputer  engineering
■650  4▼aSystems  science
■653    ▼aVisual  knowledge
■653    ▼aGenerative  models
■653    ▼aIntrinsic  images
■653    ▼aImage  decomposition
■653    ▼aDepth
■653    ▼aAlbedo
■653    ▼aSegmentation
■690    ▼a0984
■690    ▼a0464
■690    ▼a0790
■71020▼aUniversity  of  Illinois  at  Urbana-Champaign▼bComputer  Science.
■7730  ▼tDissertations  Abstracts  International▼g87-05B.
■790    ▼a0090
■791    ▼aPh.D.
■792    ▼a2024
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17361086▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

미리보기

내보내기

chatGPT토론

Ai 추천 관련 도서


    신착도서 더보기
    최근 3년간 통계입니다.

    소장정보

    • 예약
    • 소재불명신고
    • 나의폴더
    • 우선정리요청
    • 비도서대출신청
    • 야간 도서대출신청
    소장자료
    등록번호 청구기호 소장처 대출가능여부 대출정보
    TF16919 전자도서 대출가능 마이폴더 부재도서신고 비도서대출신청 야간 도서대출신청

    * 대출중인 자료에 한하여 예약이 가능합니다. 예약을 원하시면 예약버튼을 클릭하십시오.

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.