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:20260817T171532Z
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:Giuseppe Vecchio (Adobe Research, University of Catania) and R
 osalie Martin, Arthur Roullier, Adrien Kaiser, Romain Rouffet, Valentin De
 schaintre, and Tamy Boubekeur (Adobe Research)\n\nMaterial reconstruction 
 from a photograph is a key component of 3D content creation democratizatio
 n. We propose to formulate this ill-posed problem as a controlled synthesi
 s one, leveraging the recent progress in generative deep networks. We pres
 ent ControlMat, a method which, given a single photograph with uncontrolle
 d illumination as input, conditions a diffusion model to generate plausibl
 e, tileable, high-resolution physically-based digital materials. We carefu
 lly analyze the behavior of diffusion models for multi-channel outputs, ad
 apt the sampling process to fuse multi-scale information and introduce rol
 led diffusion to enable both tileability and patched diffusion for high-re
 solution outputs. Our generative approach further permits exploration of a
  variety of materials that could correspond to the input image, mitigating
  the unknown lighting conditions. We show that our approach outperforms re
 cent inference and latent-space optimization methods, and we carefully val
 idate our diffusion process design choices.\n\nRegistration Category: Full
  Access, Full Access Supporter\n\nLanguage Format: English Language\n\nSes
 sion Chair: Valentin Deschaintre (Adobe Research)\n\n
URL:https://asia.siggraph.org/2024/program/?id=tog_112&sess=sess146
END:VEVENT
END:VCALENDAR