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:20230103T035306Z
LOCATION:Auditorium\, Level 5\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T100000
DTEND;TZID=Asia/Seoul:20221206T120000
UID:siggraphasia_SIGGRAPH Asia 2022_sess153_papers_222@linklings.com
SUMMARY:Efficient Light Probes for Real-time Global Illumination
DESCRIPTION:Technical Papers\n\nEfficient Light Probes for Real-time Globa
 l Illumination\n\nGuo, Zong, Song, Fu, Tao...\n\nReproducing physically-ba
 sed global illumination (GI) effects has been a long-standing demand for m
 any real-time graphical applications. In pursuit of this goal, many recent
  engines resort to some form of light probes baked in a precomputation sta
 ge. Unfortunately, the GI effects stemming from the precomputed probes are
  rather limited due to the constraints in the probe storage, representatio
 n or query. In this paper, we propose a neural method for probe-based GI r
 endering which can generate a wide range of GI effects, including glossy r
 eflection with multiple bounces, in complex scenes. The key contributions 
 behind our work include a gradient-based search algorithm and a neural ima
 ge reconstruction method. The search algorithm is designed to reproject th
 e probes' contents to any query viewpoint, without introducing parallax er
 rors, and converges fast to the optimal solution. The neural image reconst
 ruction method, based on a dedicated neural network and several G-buffers,
  tries to recover high-quality images from low-quality inputs due to limit
 ed resolution or (potential) low sampling rate of the probes. This neural 
 method makes the generation of light probes efficient. Moreover, a tempora
 l reprojection strategy and a temporal loss are employed to improve tempor
 al stability for animation sequences. The whole pipeline runs in real-time
  (>30 frames per second) even for high-resolution (1920x1080) outputs, tha
 nks to the fast convergence rate of the gradient-based search algorithm an
 d a light-weight design of the neural network. Extensive experiments on mu
 ltiple complex scenes have been conducted to show the superiority of our m
 ethod over the state-of-the-arts.\n\nRegistration Category: FULL ACCESS, E
 XPERIENCE PLUS ACCESS, EXPERIENCE ACCESS, TRADE EXHIBITOR\n\nLanguage: ENG
 LISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_222&sess=sess15
 3
END:VEVENT
END:VCALENDAR