About bmild/nerf

The original Tensorflow reference implementation of NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis (Mildenhall et al., 2020). See the paper page for context.

This repo defines the canonical training loop for the 5D radiance field MLP, positional encoding, hierarchical volume sampling, and the LLFF / Blender synthetic dataset loaders that became standard NeRF evaluation harnesses. Nearly every later NeRF paper in the corpus (Mip-NeRF, Mip-NeRF 360, Instant-NGP, Plenoxels, NeRF-W, BARF) compares against this baseline. Mirrored here under MIT — README only.