BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:3D-based Object Understanding in The Wild
DTSTART:20180604T090000
DTEND:20180604T110000
DTSTAMP:20260502T060321Z
UID:9f4a8d7a1b5cb3e2ceab05c16f8301dbafeaa379bbab2890f8dfc8fe
CATEGORIES:Conferences - Seminars
DESCRIPTION:Edoardo Remelli\nEDIC candidacy exam\nExam president: Dr. Fran
 çois Fleuret\nThesis advisor: Prof. Pascal Fua\nCo-examiner: Prof. Martin
  Jaggi\n\nAbstract\nRecovering the 3D shape of an object from a single per
 spective has always been central to Computer Vision. Despite recent advanc
 es\, current state-of-the-art single view 3D mesh reconstruction technique
 s either do not perform well on real data or rely on strong manual input. 
 In this research proposal we first review three different approaches to 3D
  reconstruction from monocular imagery [1]\, [2]\, [3]. Then\, building up
 on them\, we propose a framework generating 3D meshes from RGB input which
 \, making use of recurrent neural networks\, can generate meshes with arbi
 trary topologies. We envision the first method able to reconstruct arbitra
 ry objects given monocular sensor data in unconstrained settings\, and we 
 are confident this will be a milestone towards general 3D-based object und
 erstanding. \n\nBackground papers\n\nWhat shape are dolphins? Building 3D
  morphable models from 2D images\, by  Cashman T.J.\,  Fitzgibbon A.\,W.
 \,\nLearning category-specific mesh reconstruction from image collections\
 , by Kanazawa A.\, et al.\n3D-PRNN: Generating shape primitives with recur
 rent neural networks\, by Zou\, C.\, et al.\n\n\n 
LOCATION:BC 329 https://plan.epfl.ch/?room==BC%20329
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR