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:20251218T030657Z
LOCATION:Meeting Room S421\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251218T090000
DTEND;TZID=Asia/Hong_Kong:20251218T091000
UID:siggraphasia_SIGGRAPH Asia 2025_sess146_papers_1209@linklings.com
SUMMARY:FreeArt3D: Training-Free Articulated Object Generation using 3D Di
 ffusion
DESCRIPTION:Chuhao Chen, Isabella Liu, Xinyue Wei, and Hao Su (University 
 of California San Diego) and Minghua Liu (Hillbot)\n\nArticulated 3D objec
 ts are central to many applications in robotics, AR/VR, and animation. Rec
 ent approaches to modeling such objects either rely on optimization-based 
 reconstruction pipelines that require dense-view supervision or on feed-fo
 rward generative models that produce coarse geometric approximations and o
 ften overlook surface texture. In contrast, open-world 3D generation of st
 atic objects has achieved remarkable success, especially with the advent o
 f native 3D diffusion models such as Trellis. However, extending these met
 hods to articulated objects by training native 3D diffusion models poses s
 ignificant challenges. In this work, we present FreeArt3D, a training-free
  framework for articulated 3D object generation. Instead of training a new
  model on limited articulated data, FreeArt3D repurposes a pre-trained sta
 tic 3D diffusion model (e.g., Trellis) as a powerful shape prior. It exten
 ds Score Distillation Sampling (SDS) into the 3D-to-4D domain by treating 
 articulation as an additional generative dimension. Given a few images cap
 tured in different articulation states, FreeArt3D jointly optimizes the ob
 ject’s geometry, texture, and articulation parameters—without requiring ta
 sk-specific training or access to large-scale articulated datasets. Our me
 thod generates high-fidelity geometry and textures, accurately predicts un
 derlying kinematic structures, and generalizes well across diverse object 
 categories. Despite following a per-instance optimization paradigm, FreeAr
 t3D completes in minutes and significantly outperforms prior state-of-the-
 art approaches in both quality and versatility.\n\nRegistration Category: 
 Full Access, Full Access Supporter\n\nSession Chair: Ziqi Wang (HKUST, EPF
 L)\n\n
END:VEVENT
END:VCALENDAR