Baking Neural Radiance Fields for Real-Time View Synthesis
This enables real-time photorealistic 3D scene rendering on commodity hardware, addressing a key bottleneck for applications like virtual reality and gaming, though it is incremental as it builds on NeRF.
The paper tackles the problem of Neural Radiance Fields (NeRF) being too slow for real-time view synthesis by introducing a method to 'bake' NeRF into a Sparse Neural Radiance Grid (SNeRG), achieving real-time rendering at over 30 frames per second on a laptop GPU with compact scene sizes averaging less than 90 MB.
Neural volumetric representations such as Neural Radiance Fields (NeRF) have emerged as a compelling technique for learning to represent 3D scenes from images with the goal of rendering photorealistic images of the scene from unobserved viewpoints. However, NeRF's computational requirements are prohibitive for real-time applications: rendering views from a trained NeRF requires querying a multilayer perceptron (MLP) hundreds of times per ray. We present a method to train a NeRF, then precompute and store (i.e. "bake") it as a novel representation called a Sparse Neural Radiance Grid (SNeRG) that enables real-time rendering on commodity hardware. To achieve this, we introduce 1) a reformulation of NeRF's architecture, and 2) a sparse voxel grid representation with learned feature vectors. The resulting scene representation retains NeRF's ability to render fine geometric details and view-dependent appearance, is compact (averaging less than 90 MB per scene), and can be rendered in real-time (higher than 30 frames per second on a laptop GPU). Actual screen captures are shown in our video.