# autonomousvision/sdfstudio

A Unified Framework for Surface Reconstruction

Repository: https://github.com/autonomousvision/sdfstudio
Canonical: https://ross.abutalabs.com/products/sdfstudio
Language: Python
License: Apache-2.0
License Family: permissive
Topics: 3d-reconstruction, implicit-neural-representation, multi-view-reconstruction, nerf, pytorch, sdf, surface-reconstruction
Last push: 2024-07-11T02:15:36+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 96
- inputs: {"age_days": 1357, "days_push": 784, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2120, forks 195 (observed 2026-08-28T04:06:16.520657+00:00)

## What it is
SDFStudio is a unified and modular framework for neural implicit surface reconstruction built on top of nerfstudio. It provides unified implementations of methods like UniSurf, VolSDF, and NeuS, along with support for various scene representations and sampling strategies.

## Use cases
- reconstruct 3d surfaces from multi-view images
- train neural implicit sdf models for scene reconstruction
- compare different neural surface reconstruction methods like neus and volsdf
- use monocular depth cues to improve 3d reconstruction
- reconstruct surfaces from rgb-d data
- experiment with different scene representations like tri-planes and feature grids
- apply monocular cues to neus for mono-neus reconstruction

## When to choose
- you need a unified framework to implement or compare neural implicit surface reconstruction methods
- you want to reconstruct high-quality 3d surfaces from multi-view or rgb-d images
- you want to build on nerfstudio's infrastructure for 3d reconstruction research
- you want to combine ideas from different reconstruction methods in a modular way

## When to avoid
- you need real-time 3d reconstruction on low-end hardware without a cuda gpu
- you only need simple point cloud registration or mesh processing rather than neural reconstruction
- you need a production-ready photogrammetry tool rather than a research framework
- you work outside python or cannot install cuda dependencies like tiny-cuda-nn

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, simulation, graphics
- domain: computer-vision, image-processing, machine-learning, graphics, autonomous-vehicles
- platform: python, cross-platform
- tags: 3d-reconstruction, implicit-neural-representation, surface-reconstruction, sdf, nerf, multi-view-reconstruction, pytorch, neural-rendering, photogrammetry, gpu, linux

## Member repositories
- autonomousvision/sdfstudio (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:16.520657+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:52:53.684233+00:00, confidence not recorded.
  - readme: https://github.com/autonomousvision/sdfstudio (fetched 2026-08-28T04:06:16.520657+00:00, sha b975cdea1143)
- Data as of 2026-08-30T08:39:29.467469+00:00.