Sakhanda Wire
NVDA $154.00 -1.37% MSFT $506.28 -2.26% GOOGL $280.87 -1.00% META $620.34 -0.97% AMZN $256.88 -0.34%
← Back to the news

Tutorial builds hierarchical NeRF pipeline with JAX3D

A tutorial presents an end-to-end hierarchical Neural Radiance Field (NeRF) pipeline for volumetric rendering, novel-view synthesis and 3D reconstruction. The workflow begins with a synthetic multi-view dataset generated from an analytic scene that combines volumetric geometry with view-dependent radiance. JAX3D primitives, including sample_along_rays and volume_rendering, define the forward rendering process.

The model uses JAX, Flax and Optax, with positional encoding, skip connections, separate coarse and fine networks, and conditioning on viewing direction. Hierarchical importance sampling is performed with sample_piecewise_constant_pdf. Training combines JAX JIT compilation, Adam optimization, exponential learning-rate decay and gradient clipping.

The tutorial also describes several evaluation and visualization steps. These include measuring novel-view synthesis with PSNR, examining depth and opacity, inspecting sampling behavior, producing 360-degree renderings and extracting geometry with marching cubes. The code sets up the required dependencies and adapts training parameters for GPU or CPU execution, while using camera poses and normalized rays to create the synthetic observations.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news