Introduction

What is SciGraphs?

Graphs are one of the most versatile abstractions in science: a set of nodes and edges encodes pairwise relationships, enabling analyses of spatial proximity, information flow, temporal evolution, or hierarchical containment. In practice, the usefulness of a graph model often depends on whether humans can perceive structure — communities, bottlenecks, symmetries, hierarchies, and anomalies that guide hypothesis formation.

Most network visualization still defaults to 2D node–link diagrams, which exhibit well-known failure modes at scale: edge crossings, overplotting, and visual “hairballs” that obscure topology when graphs become large or dense. At the same time, a persistent gap exists between analytical network exploration and high-fidelity 3D scientific communication. Many scientific problems — spatial networks (transport, migration, infrastructure), multi-layer urban contexts, or geometric/topological constraints — benefit from depth cues, lighting, and a physically meaningful environment.

SciGraphs addresses this gap. It is an open-source Blender extension that unifies graph ingestion, analysis, and rendering into a single pipeline, preserving analytical metadata throughout while providing cinematographic composition, physically based lighting, and realistic camera effects.

The two domains

SciGraphs is designed around two complementary domains.

Domain Description Typical inputs
Combinatorial graphs Spatialization derived from layout algorithms; analysis focuses on network measures and mesoscale structure. Edge lists, GEXF, GraphML, SQL queries, SuiteSparse matrices
Spatial / geometric graphs Geospatial networks and coordinate-faithful embeddings from external data sources. OSMnx street networks, Overture Maps, sparse-matrix coordinates, urban morphology

Why Blender?

Blender is a free, open-source 3D creation suite that has gained traction in scientific visualization owing to three key features:

  1. A fully scriptable Python API that exposes every aspect of the application, from scene construction to render dispatch.
  2. Two production-grade render engines — Cycles (unbiased path-tracer) and EEVEE Next (real-time rasterizer) — producing photorealistic or interactive-quality imagery.
  3. Geometry Nodes, a node-based visual programming environment for procedural geometry generation that operates on named attributes.

The Geometry Nodes attribute model maps naturally to graph analytics: metrics computed by SciGraphs are stored as per-node or per-edge named attributes, and Geometry Nodes read those attributes to drive instance scale, colour-ramp lookups, node size, or any other visual parameter — without ever modifying the source data (a non-destructive workflow). Because Geometry Nodes evaluate lazily on the GPU and use hardware instancing, a single modifier can place millions of node glyphs and edge primitives with minimal memory overhead.

NoteHow this documentation is organized
  • The Guide covers concepts, installation, the system architecture, the sidebar interface, and the end-to-end pipeline.
  • The Panel Reference documents every panel and subpanel in the sidebar, with a usage guide for each.
  • The Tutorials are step-by-step walkthroughs of each individual panel, using the Burjassot and La Paz cases.
  • The Examples are end-to-end walkthroughs of complete scenarios.
Back to top