Graph Atlas is a tool for visualizing and exploring networks of relationships. It represents elements as points, called nodes, and their relationships as lines, called edges.

For example, in a Unity game, nodes could represent NPCs and edges their relationships. In another context, they could represent cities connected by roads or ideas connected to one another. The tool is neutral: your data determines what the network means.

What is it useful for?

It helps you identify:

- The most connected elements.

- Groups with strong internal connections.

- Elements that act as bridges between groups.

- Isolated parts of a network.

- The shortest path between two elements.

Colors and node sizes make network properties easier to see. For example, coloring nodes by degree highlights those with the most connections.

How does it work?

The application performs three main operations:

1. Loads a network from a demonstration or an imported file.

1. Calculates its properties from the connections.

1. Positions and draws its nodes in an interactive scene.

Changing the layout changes how you see the network, without changing its connections. The same network can appear as a sphere, circle, helix or stack of layers.

With the “Force-directed · 3D” layout, nodes repel one another while edges pull connected nodes together, like small springs. The network gradually settles into an arrangement that helps reveal its structure.

What do the metrics mean?

Metric	What it tells you
Degree	The number of directly connected neighbors.
PageRank	A node’s importance based on its neighbors’ connections.
Betweenness	How often a node acts as a bridge along shortest paths.
Harmonic closeness	How close a node is to others, measured in edge steps.
Community	A group suggested by the algorithm based on connections.
Connected component	A group in which every node can reach every other node through edges.

A detected community is a structural indication, not proof that its members share a particular real-world relationship.

◈ Graph Atlas

Network laboratory

01 / Network

Demo scene

Seed

Find a node

02 / Layout

Node layout Scale 100 %

Auto-rotate

2D view

03 / Appearance

Color by Node size Edge opacity Edge style Minimum degree 0

Show labels

EXPLORATION / 3D

Community constellation

Phase 100 %

From SciGraphs to the browser

This adaptation recreates the visualization principles and several motifs from the demonstration. It is not a complete port of the Blender software.

Analysis of the 54-second video

The montage alternates between branching structures, a spherical network, colored clusters, a mesh, bundled edges in a ring, orbital paths, a double helix and a matrix representation. The black background isolates the shapes; colors distinguish groups; camera movement and transformations reveal their structure. These scenes are procedural reconstructions, not the original datasets.

Feature comparison

Feature	This HTML adaptation
Import and export	Import JSON, CSV, GEXF, GraphML and Matrix Market; export JSON, CSV, PNG and standalone HTML.
Analysis	Degree, PageRank, Brandes betweenness, harmonic closeness, components, label-propagation communities and BFS shortest paths.
Layout and animation	3D forces, sphere, circle, helix and layers; interpolation, rotation and a cycling demonstration.
Rendering and interaction	3D projection in Canvas, depth, selection, search, filters, curves and center bundles.
Out of scope	Cycles/EEVEE, Geometry Nodes, the SciGraphs GPU engine, millions of elements, OSM/City2Graph, SQL, GTFS, advanced topology and the extension’s complete algorithm collection.

The original project combines graph processing, visual attributes and rendering in Blender. Its computation libraries are separate from the Blender integration. Geospatial features and native rendering would require additional components to achieve full equivalence.

Sources: SciGraphs repository · documentation · José Marín Fariña’s video.

Import your data

Files stay in your browser. Limit: 600 nodes and 8,000 edges to keep analysis interactive. Duplicate edges and self-loops are removed. Directed graphs and weights other than 1 are rejected to prevent incorrect interpretation.

JSON : {"nodes":[{"id":"A","label":"Alpha","x":0,"y":0,"z":0},{"id":"B"}],"edges":[{"source":"A","target":"B"}]}. CSV: header source,target ; quoted fields are supported. Positions are optional. Matrix Market: a square matrix in coordinate format; nonzero entries must equal 1. General matrices must be symmetric.

Communities use a deterministic heuristic, not SciGraphs’ SurpriseMe methods. Bundles are geometric styling, not analytical hierarchical bundling. Phase interpolates between a sphere and the chosen layout; it is not data history. Spatial data is centered and scaled for display, without geographic reprojection. JSON exports these scene coordinates. PageRank uses iterative computation with a convergence tolerance.

Graph Atlas