# Linketic/CityGaussian

[ECCV`24&ICLR`25] CityGaussian Series for High-quality Large-Scale Scene Reconstruction with Gaussians

Repository: https://github.com/Linketic/CityGaussian
Canonical: https://ross.abutalabs.com/products/citygaussian
Homepage: https://dekuliutesla.github.io/CityGaussianV2
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Topics: 3d, computer-vision, eccv2024, gaussian-splatting, graphics, neural-network, neural-rendering, novel-view-synthesis, radiance-field, level-of-details, surface-reconstruction, iclr2025, large-scale
Last push: 2026-08-16T13:02:11+00:00

## Health v2 (maintenance only)
Score: 66/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 28, longevity 61
- inputs: {"age_days": 859, "days_push": 17, "days_rel": 588, "gap_med": 31, "n_releases_24m": 4}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1251, forks 110 (observed 2026-08-28T04:04:08.266228+00:00)

## What it is
Official implementation of the CityGaussian series (ECCV 2024, ICLR 2025) for high-quality large-scale 3D scene reconstruction with Gaussian Splatting. It provides multi-GPU training, mesh extraction, trajectory-aligned rendering, and joint pose optimization built on Gaussian Lightning.

## Use cases
- reconstruct large-scale city scenes from multi-view images with gaussian splatting
- extract meshes from 3dgs reconstructions of large scenes
- render novel view videos along camera trajectories
- train 3dgs across multiple gpus with bounded memory
- refine imperfect COLMAP poses jointly with 3dgs optimization
- evaluate geometric accuracy of large-scale scene reconstructions

## When to choose
- you need real-time high-quality rendering of city-scale or other large scenes
- you want geometrically accurate surfaces and mesh extraction from gaussian splatting
- you need multi-GPU reconstruction with controllable memory cost
- you want a research-grade 3DGS pipeline with pose refinement and web viewer

## When to avoid
- you need a permissively licensed library for commercial use (CC BY-NC-SA 4.0)
- you only need small-object or single-room 3DGS reconstruction without large-scale partitioning
- you need a turnkey photogrammetry tool rather than a research codebase
- you lack GPU resources, as training is GPU-intensive

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, graphics, machine-learning, image-processing
- domain: computer-vision, graphics, artificial-intelligence, deep-learning
- platform: python
- tags: gaussian-splatting, 3dgs, novel-view-synthesis, radiance-field, mesh-extraction, large-scale-scenes, research-code, eccv2024, iclr2025, multi-gpu-training, gpu, linux, cuda

## Member repositories
- Linketic/CityGaussian (main) score 66

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:08.266228+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 2026-08-30T05:07:36.970477+00:00, confidence not recorded.
  - readme: https://github.com/Linketic/CityGaussian (fetched 2026-08-28T04:04:08.266228+00:00, sha f6b368f82f5f)
  - homepage: https://dekuliutesla.github.io/CityGaussianV2 (fetched 2026-08-29T12:18:33.052843+00:00, sha f3c7fea5094f)
- Data as of 2026-08-30T08:39:29.467469+00:00.