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:20251215T135300
DTEND;TZID=Asia/Hong_Kong:20251215T140400
UID:siggraphasia_SIGGRAPH Asia 2025_sess106_tog_121@linklings.com
SUMMARY:GS-ROR2: Bidirectional-guided 3DGS and SDF for Reflective Object  
 Relighting and Reconstruction
DESCRIPTION:Zuo-Liang Zhu (Nankai University), Beibei Wang (Nanjing Univer
 sity), and Jian Yang (Nankai University)\n\n3D Gaussian Splatting (3DGS) h
 as shown a powerful capability for novel view synthesis due to its detaile
 d expressive ability and highly efficient rendering speed. Unfortunately, 
 creating relightable 3D assets and reconstructing faithful geometry with 3
 DGS is still problematic, particularly for reflective objects, as its disc
 ontinuous representation raises difficulties in constraining geometries. I
 n contrary, volumetric signed distance field (SDF) methods provide robust 
 geometry reconstruction, while the expensive ray marching hinders its real
 -time application and slows the training. Besides, these methods struggle 
 to capture sharp geometric details. To this end, we\n propose to guide 3DG
 S and SDF bidirectionally in a complementary manner, including an SDF-aide
 d Gaussian splatting for efficient optimization of the relighting model an
 d a GS-guided SDFenhancementforhigh-quality geometry reconstruction. At th
 e core of our SDF-aided Gaussian splatting is the mutual supervision of th
 e depth and normal between blended Gaussians and SDF,\n which avoids the e
 xpensive volume rendering of SDF. Thanks to this mutual supervision, the l
 earned blended Gaussians are well-constrained with a minimal time cost. As
  the Gaussians are rendered in a deferred shading mode, the alpha-blended 
 Gaussians are smooth, while individual Gaussians may still be outliers, yi
 elding floater artifacts. Therefore, we introduce an SDF-aware  pruning st
 rategy to remove Gaussian outliers located distant from the surface define
 d by SDF, avoiding the floater issue. This way, our GS framework provides 
 reasonable normal and achieves realistic relighting, while the mesh of tru
 ncated SDF (TSDF) fusion from depth is still problematic. Therefore, we de
 sign a GS-guided SDF refinement, which utilizes the blended normal from  G
 aussians to finetune SDF. Equipped with the efficient enhancement, our met
 hod can further provide high-quality meshes for reflective objects at the 
  cost of 17% extra training time.\n\nRegistration Category: Full Access, F
 ull Access Supporter\n\nSession Chair: Yuchi Huo (Zhejiang University, Kor
 ea Advanced Institute of Science and Technology)\n\n
END:VEVENT
END:VCALENDAR