BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250110T023313Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T133400
DTEND;TZID=Asia/Tokyo:20241206T134600
UID:siggraphasia_SIGGRAPH Asia 2024_sess146_tog_112@linklings.com
SUMMARY:ControlMat: A Controlled Generative Approach to Material Capture
DESCRIPTION:Technical Papers\n\nGiuseppe Vecchio (Adobe Research, Universi
 ty of Catania) and Rosalie Martin, Arthur Roullier, Adrien Kaiser, Romain 
 Rouffet, Valentin Deschaintre, and Tamy Boubekeur (Adobe Research)\n\nMate
 rial reconstruction from a photograph is a key component of 3D content cre
 ation democratization. We propose to formulate this ill-posed problem as a
  controlled synthesis one, leveraging the recent progress in generative de
 ep networks. We present ControlMat, a method which, given a single photogr
 aph with uncontrolled illumination as input, conditions a diffusion model 
 to generate plausible, tileable, high-resolution physically-based digital 
 materials. We carefully analyze the behavior of diffusion models for multi
 -channel outputs, adapt the sampling process to fuse multi-scale informati
 on and introduce rolled diffusion to enable both tileability and patched d
 iffusion for high-resolution outputs. Our generative approach further perm
 its exploration of a variety of materials that could correspond to the inp
 ut image, mitigating the unknown lighting conditions. We show that our app
 roach outperforms recent inference and latent-space optimization methods, 
 and we carefully validate our diffusion process design choices.\n\nRegistr
 ation Category: Full Access, Full Access Supporter\n\nLanguage Format: Eng
 lish Language\n\nSession Chair: Valentin Deschaintre (Adobe Research)
URL:https://asia.siggraph.org/2024/program/?id=tog_112&sess=sess146
END:VEVENT
END:VCALENDAR