NeRF

TLDR

The paper proposes a method that achieves state‑of‑the‑art novel‑view synthesis of complex scenes by optimizing a continuous volumetric scene function from a sparse set of input views. The algorithm represents a scene with a fully connected deep network mapping 5‑D coordinates (x, y, z, θ, ϕ) to volume density and view‑dependent radiance, queries these coordinates along camera rays, and uses differentiable volume rendering to synthesize images from known camera poses. The method produces photorealistic novel views that outperform prior neural rendering and view‑synthesis approaches on complex scenes.

Abstract

We present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. Our algorithm represents a scene using a fully connected (nonconvolutional) deep network, whose input is a single continuous 5D coordinate (spatial location ( x , y , z ) and viewing direction ( θ, ϕ )) and whose output is the volume density and view-dependent emitted radiance at that spatial location. We synthesize views by querying 5D coordinates along camera rays and use classic volume rendering techniques to project the output colors and densities into an image. Because volume rendering is naturally differentiable, the only input required to optimize our representation is a set of images with known camera poses. We describe how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrate results that outperform prior work on neural rendering and view synthesis.