BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240214T070245Z
LOCATION:Meeting Room C4.11\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231213T165000
DTEND;TZID=Australia/Melbourne:20231213T170000
UID:siggraphasia_SIGGRAPH Asia 2023_sess153_papers_661@linklings.com
SUMMARY:VET: Visual Error Tomography for Point Cloud Completion and High-Q
 uality Neural Rendering
DESCRIPTION:Technical Communications, Technical Papers\n\nLinus Franke, Da
 rius Rückert, and Laura Fink (Friedrich-Alexander Universität Erlangen-Nür
 nberg); Matthias Innmann (NavVis GmbH); and Marc Stamminger (Friedrich-Ale
 xander Universität Erlangen-Nürnberg)\n\nIn the last few years, deep neura
 l networks opened the doors for big advances in novel view synthesis. Many
  of these approaches are based on a (coarse) proxy geometry obtained by st
 ructure from motion algorithms. Small deficiencies in this proxy can be fi
 xed by neural rendering, but larger holes or missing parts, as they common
 ly appear for thin structures or for glossy regions however still lead to 
 very distracting artifacts and temporal instability. In this paper, we pre
 sent a novel neural rendering based approach to detect and fix such defici
 encies. As a proxy, we use a point cloud, which allows us to easily remove
  outlier geometry and to fill in missing geometry without complicated topo
 logical operations. Keys to our approach are (i) a differentiable, blendin
 g point-based renderer that can blend out redundant points, as well as (ii
 ) the concept of Visual Error Tomography (VET), which allows us to lift 2D
  error maps to identify 3D-regions lacking geometry and to spawn novel poi
 nts accordingly. Furthermore, (iii) by adding points as nested environment
  maps, our approach allows us to generate high-quality renderings of the s
 urroundings in the same pipeline. In our results, we show that our approac
 h can significantly improve the quality of a point cloud obtained by struc
 ture from motion and thus increase novel view synthesis quality. In contra
 st to point growing techniques, the approach can also fix large-scale hole
 s and missing thin structures effectively. Rendering quality outperforms s
 tate-of-the-art methods and temporal stability is significantly improved, 
 while rendering is possible with real-time frame rates.\n\nRegistration Ca
 tegory: Full Access\n\nSession Chair: Jonah Brucker-Cohen (Lehman College 
 / CUNY, New Inc.)
URL:https://asia.siggraph.org/2023/full-program?id=papers_661&sess=sess153
END:VEVENT
END:VCALENDAR