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:20240214T070241Z
LOCATION:Darling Harbour Theatre\, Level 2 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T093000
DTEND;TZID=Australia/Melbourne:20231212T124500
UID:siggraphasia_SIGGRAPH Asia 2023_sess209_papers_902@linklings.com
SUMMARY:FLARE: Fast Learning of Animatable and Relightable Mesh Avatars
DESCRIPTION:Technical Papers\n\nShrisha Bharadwaj (Max Planck Institute fo
 r Intelligent Systems); Yufeng Zheng (ETH Zürich, Max Planck Institute for
  Intelligent Systems); Otmar Hilliges (ETH Zürich); and Michael Black and 
 Victoria Fernandez Abrevaya (Max Planck Institute for Intelligent Systems)
 \n\nOur goal is to efficiently learn personalized animatable 3D head avata
 rs from videos that are geometrically accurate, realistic, relightable, an
 d compatible with current rendering systems. While 3D meshes enable effici
 ent processing and are highly portable, they lack realism in terms of shap
 e and appearance. Neural representations, on the other hand, are realistic
  but lack compatibility and are slow to train and render. Our key insight 
 is that it is possible to efficiently learn high-fidelity 3D mesh represen
 tations via differentiable rendering, by exploiting highly-optimized metho
 ds from traditional computer graphics and approximating some of the compon
 ents with neural networks. Specifically, we introduce FLARE, a technique t
 hat enables fast creation of animatable and relightable mesh avatars from 
 a single monocular video. First, we learn a canonical geometry using a mes
 h representation, enabling efficient differentiable rasterization and stra
 ightforward animation via learned blendshapes and linear blend skinning we
 ights. Second, we follow physically-based rendering and factor observed co
 lors into intrinsic albedo, roughness, and a neural representation of the 
 illumination, allowing the learned avatars to be relit in novel scenes. Si
 nce our input videos are captured on a single device with a narrow field o
 f view, modeling the surrounding environment light is non-trivial. Based o
 n the split-sum approximation for modeling specular reflections, we addres
 s this by approximating the pre-filtered environment map with a multi-laye
 r perceptron (MLP) modulated by the surface roughness, eliminating the nee
 d to explicitly model the light. We demonstrate that our mesh-based avatar
  formulation, combined with learned deformation, material and lighting MLP
 s, produces avatars with high-quality geometry and appearance, while also 
 being efficient to train and render compared to existing approaches.\n\nRe
 gistration Category: Full Access, Enhanced Access, Trade Exhibitor, Experi
 ence Hall Exhibitor
URL:https://asia.siggraph.org/2023/full-program?id=papers_902&sess=sess209
END:VEVENT
END:VCALENDAR