BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Meshing Neural Unsigned Distance Fields
DTSTART:20220712T103000
DTEND:20220712T123000
DTSTAMP:20260503T142915Z
UID:e9516e396f969c3ca96942ad5dbbf260c8d4a23e74a449740ed3929d
CATEGORIES:Conferences - Seminars
DESCRIPTION:Federico Stella\nEDIC candidacy exam\nExam president: Prof. Am
 ir Zamir\nThesis advisor: Prof. Pascal Fua\nCo-examiner: Prof. Nicolas Fla
 mmarion\n\nAbstract\nReconstructing shapes from noisy or partial input dat
 a is a fundamental problem with several applications in different fields s
 uch as robotics\, graphics\, medicine\, architecture and virtual reality. 
 Recently\, implicit representations parametrized by neural networks and ba
 sed on Signed Distance Fields or on Occupancy Fields have emerged as a ver
 y powerful tool to represent\, manipulate and generate watertight shapes w
 ithout being constrained by resolution or by a fixed topology. Explicit me
 shes can be obtained from them by using isosurface extraction algorithms s
 uch as the popular Marching Cubes. However\, a major limitation of the exi
 sting approaches based on implicit representations is that they can only w
 ork with watertight surfaces with a clear concept of inside and outside\, 
 whereas real world data usually comes from 3D scans and needs to be pre-pr
 ocessed in order to be made watertight\, losing the inner details in the p
 rocess\, for example the inside of a car. Moreover\, 3D shapes like garmen
 ts could be more naturally parametrized as open surfaces\, allowing their 
 use in applications that require single-sided surfaces\, such as draping s
 imulators. For these reasons implicit representations based on Unsigned Di
 stance Fields have emerged\, for which classical meshing algorithms do not
  directly work\, posing additional challenges in obtaining open meshes as 
 outputs.\nIn this work\, we select three papers on the topics of geometry 
 learning\, neural unsigned distance fields approximation and isosurface ex
 traction\, highlighting the strengths and shortcomings of current methods 
 and identifying interesting research directions to generate high-quality o
 pen meshes from learned (and thus imperfect) Unsigned Distance Fields.\n\n
 Background papers\n- "A survey on deep geometry learning: From a represent
 ation perspective” by Yun-Peng Xiao\, Yu-Kun Lai\, Fang-Lue Zhang\, Ch
 unpeng Li and Lin Gao. https://arxiv.org/abs/2002.07995\n\n- "Neural Uns
 igned Distance Fields for Implicit Function Learning” by Julian Chibane\
 , Aymen Mir and Gerard Pons-Moll. https://arxiv.org/abs/2010.13938\n\n-
  “Neural Dual Contouring” by Zhiqin Chen\, A. Tagliasacchi\, T. Funk
 houser and Hao Zhang. https://arxiv.org/abs/2202.01999
LOCATION:BC 329 https://plan.epfl.ch/?room==BC%20329
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR