Copyright © 2026 OyList. All Rights Reserved.
Share
Items26
25
Neural radiance field
NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis, ECCV 2020
Ben Mildenhall, Pratul Srinivasan, Matthew Tancik, Jonathan Barron, Ravi Ramamoorthi and Ren Ng skipped building a 3D model altogether. From a set of ordinary photographs, they trained a fully connected neural network that takes a position in space and a viewing direction, and they rendered new views by querying it along camera rays.
The leap is capturing reality rather than modeling it: fine detail and view-dependent shine come out of photographs with no hand-built geometry. The paper received a Best Paper Honorable Mention, and radiance fields became one of the busiest research topics in graphics.
More about Neural radiance field26
Gaussian splatting
3D Gaussian Splatting for Real-Time Radiance Field Rendering, SIGGRAPH 2023
NeRF captured photographic scenes but could not show complete, unbounded ones at real-time rates. Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler and George Drettakis represented the scene with 3D Gaussians instead, soft colored blobs fitted to the photographs, which avoids wasting computation on empty space.
Their target was high-quality new views at 1080p and at least 100 frames per second. That brought photographic capture into the same real-time world as games, and it is the newest entry here because it changes what a captured scene can be used for.
More about Gaussian splatting