BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035311Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221208T153000
DTEND;TZID=Asia/Seoul:20221208T170000
UID:siggraphasia_SIGGRAPH Asia 2022_sess166_papers_588@linklings.com
SUMMARY:Neural Photo-Finishing
DESCRIPTION:Technical Communications, Technical Papers\n\nNeural Photo-Fin
 ishing\n\nTseng, Zhang, Jebe, Zhang, Xia...\n\nImage processing pipelines 
 are ubiquitous and we rely on them either directly, by filtering or adjust
 ing an image post-capture, or indirectly, as image signal processing pipel
 ine (ISP) on broadly deployed camera systems. Used by artists, photographe
 rs, system engineers, and for downstream vision tasks, traditional image p
 rocessing pipelines feature complex algorithmic branches developed over de
 cades. Recently, image-to-image networks have made great strides in image 
 processing, style transfer, and semantic understanding. The differentiable
  nature of these networks allows them to fit a large corpus of data, howev
 er, they do not allow for intuitive, fine-grained controls that photograph
 ers find in modern photo-finishing tools.\n\nThis work closes this gap and
  presents an approach to making complex photo-finishing pipelines differen
 tiable, allowing legacy algorithms to be trained akin to neural networks u
 sing first-order optimization methods. By concatenating tailored network p
 roxy models of individual processing steps (e.g. white-balance, tone-mappi
 ng, color tuning), we can model a non-differentiable reference image finis
 hing pipeline more faithfully than existing proxy image-to-image network m
 odels. We validate the method for several diverse applications, including 
 photo and video style transfer, slider regression for commercial camera IS
 Ps, photography-driven neural demosaicking, and adversarial photo-editing.
 \n\nRegistration Category: FULL ACCESS, ON-DEMAND ACCESS\n\nLanguage: ENGL
 ISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_588&sess=sess16
 6
END:VEVENT
END:VCALENDAR