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 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241204T172800
DTEND;TZID=Asia/Tokyo:20241204T174000
UID:siggraphasia_SIGGRAPH Asia 2024_sess121_papers_946@linklings.com
SUMMARY:FragmentDiff: A Diffusion Model for Fractured Object Assembly
DESCRIPTION:Qun-Ce Xu and Hao-Xiang Chen (BNRist, Department of Computer S
 cience and Technology, Tsinghua University); Jiacheng Hua (Department of C
 omputer Science and Technology, Tsinghua University); Xiaohua Zhan (Depart
 ment of Foreign Languages and Literatures, Tsinghua University); Yong-Lian
 g Yang (Department of Computer Science, University of Bath); and Tai-Jiang
  Mu (BNRist, Department of Computer Science and Technology, Tsinghua Unive
 rsity)\n\nFractured object reassembly is a challenging problem in computer
  vision and graphics with applications in industrial manufacturing and arc
 haeology. Traditional methods based on shape descriptors and geometric reg
 istration often struggle with ambiguous features, resulting in lower accur
 acy. To address this, we propose a novel approach inspired by diffusion mo
 dels and 3D transformers. Our method applies diffusion denoising combined 
 with a 3D transformer to predict the pose parameter of each fragment. We e
 valuate our approach on a fractured object dataset and demonstrate superio
 r performance compared to state-of-the-art methods. Our method offers a pr
 omising solution for accurate and robust fractured object reassembly, adva
 ncing the field of computer vision in complex shape analysis and assembly 
 tasks.\n\nRegistration Category: Full Access, Full Access Supporter\n\nLan
 guage Format: English Language\n\nSession Chair: Nobuyuki Umetani (Univers
 ity of Tokyo)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_946&sess=sess121
END:VEVENT
END:VCALENDAR