BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030653Z
LOCATION:Meeting Room S426+S427\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251216T132000
DTEND;TZID=Asia/Hong_Kong:20251216T133100
UID:siggraphasia_SIGGRAPH Asia 2025_sess122_tog_102@linklings.com
SUMMARY:Many-Worlds Inverse Rendering
DESCRIPTION:Ziyi Zhang, Nicolas Roussel, and Wenzel Jakob (Ecole Polytechn
 ique Fédérale de Lausanne)\n\nDiscontinuous visibility changes remain a ma
 jor bottleneck when optimizing surfaces within a physically based inverse 
 renderer. Many previous works have proposed sophisticated algorithms and d
 ata structures to sample visibility silhouettes more efficiently.\n\nOur w
 ork presents another solution: instead of differentiating a tentative surf
 ace locally, we differentiate a non-local perturbation of a surface. We re
 fer to this as a “many-worlds” representation because it models a non-inte
 racting superposition of conflicting explanations (“worlds”) of the input 
 dataset. Each world is optically isolated from others, leading to a new tr
 ansport law that distinguishes our method from prior work based on exponen
 tial random media.\n\nThe resulting Monte Carlo algorithm is simpler and m
 ore efficient than prior methods. We demonstrate that our method promotes 
 rapid convergence, both in terms of the total iteration count and the cost
  per iteration.\n\nRegistration Category: Full Access, Full Access Support
 er\n\nSession Chair: Elena Garces (Adobe, Universidad Rey Juan Carlos)\n\n
END:VEVENT
END:VCALENDAR