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:20241205T165300
DTEND;TZID=Asia/Tokyo:20241205T170500
UID:siggraphasia_SIGGRAPH Asia 2024_sess137_papers_887@linklings.com
SUMMARY:StableNormal: Reducing Diffusion Variance for Stable and Sharp Nor
 mal
DESCRIPTION:Chongjie Ye and Lingteng Qiu (FNii, The Chinese University of 
 Hong Kong, Shenzhen; SSE, The Chinese University of Hong Kong, Shenzhen); 
 Xiaodong Gu and Qi Zuo (Alibaba); Yushuang Wu (FNii, The Chinese Universit
 y of Hong Kong, Shenzhen; SSE, The Chinese University of Hong Kong, Shenzh
 en); Zilong Dong and Liefeng Bo (Alibaba); Yuliang Xiu (Max Planck Institu
 te for Intelligent Systems); and Xiaoguang Han (SSE, The Chinese Universit
 y of Hong Kong, Shenzhen; FNii, The Chinese University of Hong Kong, Shenz
 hen)\n\nThis work addresses the challenge of high-quality surface normal e
 stimation from monocular colored inputs (i.e., images and videos), a field
  which has recently been revolutionized by repurposing diffusion priors. H
 owever, previous attempts still struggle with stochastic inference, confli
 cting with the deterministic nature of the Image2Normal task, and costly e
 nsembling step, which slows down the estimation process. Our method, Stabl
 eNormal, mitigates the stochasticity of the diffusion process by reducing 
 inference variance, thus producing “Stable-and-Sharp” normal estimates wit
 hout any additional ensembling process. StableNormal works robustly under 
 chal lenging imaging conditions, such as extreme lighting, blurring, and l
 ow quality. It is also robust against transparent and reflective surfaces,
  as well as cluttered scenes with numerous objects. Specifically, StableNo
 rmal employs a coarse-to-fine strategy, which starts with a one-step norma
 l estimator (YOSO) to derive an initial normal guess, that is relatively c
 oarse but reliable, then followed by a semantic-guided refinement process 
 (SG-DRN) that refines the normals to recover geometric details. The effect
 iveness of StableNormal is demonstrated through competitive performance in
  standard datasets such as DIODE-indoor, iBims, ScannetV2, and NYUv2, and 
 also in various downstream tasks, such as surface reconstruction and norma
 l enhancement. These results evidence that StableNormal retains both the “
 stability” and “sharpness” for accurate normal estimation. StableNormal re
 presents a baby attempt to repurpose diffusion priors for deterministic es
 timation. To democratize this, code and models have been publicly availabl
 e in hf.co/Stable-X.\n\nRegistration Category: Full Access, Full Access Su
 pporter\n\nLanguage Format: English Language\n\nSession Chair: Michael Rub
 instein (Google)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_887&sess=sess137
END:VEVENT
END:VCALENDAR