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:20251218T030656Z
LOCATION:Meeting Room S426+S427\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251217T151100
DTEND;TZID=Asia/Hong_Kong:20251217T152200
UID:siggraphasia_SIGGRAPH Asia 2025_sess142_papers_1712@linklings.com
SUMMARY:AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Vi
 ews
DESCRIPTION:Lihan Jiang (University of Science and Technology of China, Sh
 anghai Artificial Intelligence Laboratory); Yucheng Mao (Shanghai Artifici
 al Intelligence Laboratory); Linning Xu (Chinese University of Hong Kong);
  Tao Lu (Brown University); Kerui Ren (Shanghai Jiao Tong University, Shan
 ghai Artificial Intelligence Laboratory); Yichen Jin, Xudong Xu, Mulin Yu,
  and Jiangmiao Pang (Shanghai Artificial Intelligence Laboratory); Feng Zh
 ao (University of Science and Technology of China); Dahua Lin (Chinese Uni
 versity of Hong Kong); and Bo Dai (University of Hong Kong)\n\nWe introduc
 e AnySplat, a feed‑forward network for novel‑view synthesis from uncalibra
 ted image collections. In contrast to traditional neural‑rendering pipelin
 es that demand known camera poses and per‑scene optimization, or recent fe
 ed‑forward methods that buckle under the computational weight of dense vie
 ws—our model predicts everything in one shot. A single forward pass yields
  (1) a set of 3D Gaussian primitives encoding both scene geometry and appe
 arance, and (2) the corresponding camera intrinsics and extrinsics for eac
 h input image. This unified design scales effortlessly to casually capture
 d, multi‑view datasets without any pose annotations. In extensive zero‑sho
 t evaluations, AnySplat matches the quality of pose‑aware baselines in bot
 h sparse‑ and dense‑view scenarios while surpassing existing pose‑free app
 roaches. Moreover, it greatly reduce rendering latency compared to optimiz
 ation‑based neural fields, bringing real‑time novel‑view synthesis within 
 reach for unconstrained capture settings. Project page: https://city-super
 .github.io/anysplat/.\n\nRegistration Category: Full Access, Full Access S
 upporter\n\nSession Chair: Pedro Sander (HKUST)\n\n
END:VEVENT
END:VCALENDAR