BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Brisbane
X-LIC-LOCATION:Australia/Brisbane
BEGIN:STANDARD
DTSTART:19920301T030000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20191203T080637Z
LOCATION:Great Hall 1&2
DTSTART;TZID=Australia/Brisbane:20191117T192900
DTEND;TZID=Australia/Brisbane:20191117T193000
UID:siggraphasia_SIGGRAPH Asia 2019_sess222_papers_347@linklings.com
SUMMARY:Biomimetic Eye Modeling & Deep Neuromuscular Oculomotor Control
DESCRIPTION:Technical Papers Fast-Forward \n\nBiomimetic Eye Modeling & De
 ep Neuromuscular Oculomotor Control\n\nNakada, Lakshmipathy, Zhou, Chen, L
 ing...\n\nWe present a novel, biomimetic model of the eye for realistic vi
 rtual\nhuman animation. We also introduce a deep learning approach to\nocu
 lomotor control that is compatible with our biomechanical eye\nmodel. Our 
 eye model consists of the following functional components:\n(i) submodels 
 of the 6 extraocular muscles that actuate realistic eye\nmovements, (ii) a
 n iris submodel, actuated by pupillary muscles, that\naccommodates to inco
 ming light intensity, (iii) a corneal submodel and\na deformable, ciliary-
 muscle-actuated lens submodel, which refract\nincoming light rays for foca
 l accommodation, and (iv) a retina with a\nmultitude of photoreceptors arr
 anged in a biomimetic, foveated\ndistribution. The light intensity capture
 d by the photoreceptors is\ncomputed using ray tracing from the photorecep
 tor positions through\nthe finite aperture pupil into the 3D virtual envir
 onment, and the\nvisual information is output by the eye via an optic nerv
 e vector. Our\noculomotor control system includes a foveation controller i
 mplemented\nas a locally-connected, irregular Deep Neural Network (DNN) th
 at\nconforms to the nonuniform retinal photoreceptor distribution, and a\n
 neuromuscular motor controller implemented as a fully-connected DNN,\nplus
  auxiliary Shallow Neural Networks (SNNs) that control the\naccommodation 
 of the pupil and lens. The DNNs are trained offline\nthrough deep learning
  from data synthesized by the eye model itself.\nOnce trained, the oculomo
 tor control system operates robustly and\nefficiently online. It innervate
 s the intraocular muscles to perform\nillumination and focal accommodation
  and the extraocular muscles to\nproduce natural eye movements in order to
  foveate and pursue moving\nvisual targets. We additionally demonstrate th
 e operation of our eye\nmodel (binocularly) within a recently introduced s
 ensorimotor control\nframework involving an anatomically-accurate biomecha
 nical human\nmusculoskeletal model.\n\nRegistration Category: Platinum Pas
 s, Full Conference Pass, Full Conference One-Day Pass, Basic Conference Pa
 ss, Student One-Day Pass, Exhibitor Pass
URL:https://sa2019.conference-program.com/presentation?id=papers_347&sess=
 sess222
END:VEVENT
END:VCALENDAR

