Presentation

· Contributors · Organizations · Search Program · My Schedule · Happening Now · Maps

SMERF: Streamable Memory Efficient Radiance Fields for Real-time Large-scene Exploration

DescriptionSMERF is a method for real-time view-synthesis of large, multi-room spaces on resource-constrained devices. By distilling state-of-the-art NeRF models into a compact, hierarchical representation, SMERF enables the exploration of spaces as large as 300 m^2 with quality approaching that of its teacher while rendering three orders of magnitude faster.

Authors

Tübingen AI Center, University of Tübingen

Peter Zhizhin

Google Research

Jean-François Thibert

Google AR/VR