BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:Asia/Seoul X-LIC-LOCATION:Asia/Seoul BEGIN:STANDARD TZOFFSETFROM:+0900 TZOFFSETTO:+0900 TZNAME:KST DTSTART:18871231T000000 DTSTART:19881009T020000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20230103T035307Z LOCATION:Room 324\, Level 3\, West Wing DTSTART;TZID=Asia/Seoul:20221206T153000 DTEND;TZID=Asia/Seoul:20221206T170000 UID:siggraphasia_SIGGRAPH Asia 2022_sess155_papers_322@linklings.com SUMMARY:Differentiable Rendering of Neural SDFs through Reparameterization DESCRIPTION:Technical Communications, Technical Papers\n\nDifferentiable R endering of Neural SDFs through Reparameterization\n\nBangaru, Gharbi, Lua n, Li, Sunkavalli...\n\nWe present a method to automatically compute corre ct gradients with respect to geometric scene parameters in neural SDF rend erers. Recent physically-based differentiable rendering techniques for mes hes have used edge-sampling to handle discontinuities, particularly at obj ect silhouettes, but SDFs do not have a simple parametric form amenable to sampling. Instead, our approach builds on area-sampling techniques and de velops a continuous warping function for SDFs to account for these discont inuities. Our method leverages the distance to surface encoded in an SDF a nd uses quadrature on sphere tracer points to compute this warping functio n. We further show that this can be done by subsampling the points to make the method tractable for neural SDFs. Our differentiable renderer can be used to optimize neural shapes from multi-view images and produces compara ble 3D reconstructions to recent SDF-based inverse rendering methods, with out the need for 2D segmentation masks to guide the geometry optimization and no volumetric approximations to the geometry.\n\nRegistration Category : FULL ACCESS, ON-DEMAND ACCESS\n\nLanguage: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND URL:https://sa2022.siggraph.org/en/full-program/?id=papers_322&sess=sess15 5 END:VEVENT END:VCALENDAR