sjtuytc/UnboundedNeRFPytorch

Health v2 · maintenance only

Adoption not part of the score

1324 stars · 111 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A PyTorch implementation benchmarking state-of-the-art unbounded (large-scale) neural radiance field methods like NeRF++, DVGO, and Block-NeRF. It provides a simple, fast codebase achieving SOTA PSNR on Mip-NeRF-360 and Tanks & Temples benchmarks.

Use cases

• train unbounded NeRF models on large-scale scenes
• reproduce Mip-NeRF-360 benchmark results
• render novel views of outdoor 360-degree scenes
• benchmark radiance field algorithms like NeRF++ and DVGO
• train Block-NeRF style models on street-level imagery
• research large-scale neural radiance fields in PyTorch

When to choose

• you need a simple, fast PyTorch codebase for unbounded/large-scale NeRFs
• you want SOTA-quality novel view synthesis on 360-degree outdoor scenes
• you want to benchmark multiple radiance field methods in one repo

When to avoid

• you need a production-ready or well-supported library rather than a research project in progress
• you need bounded/indoor-only NeRF with a polished API
• you lack a GPU or PyTorch environment

Facets

library · maturity active

machine-learning deep-learning graphics simulation computer-vision graphics deep-learning python nerf neural-radiance-fields view-synthesis pytorch block-nerf mip-nerf-360 unbounded-scenes research-code research gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="sjtuytc/UnboundedNeRFPytorch")