NeRF
Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik et al. · Communications of the ACM · 2021
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.
Cited by 6,744 works. doi:10.1145/3503250 Open-access PDF
Authors
References
- A volumetric method for building complex models from range images
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
- Modeling and rendering architecture from photographs
- Structure-from-Motion Revisited
- NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
- Ray tracing volume densities
- DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation
- Unstructured lumigraph rendering
- Occupancy Networks: Learning 3D Reconstruction in Function Space
- Optical models for direct volume rendering