Vault Analyser Documentation

Vault Analyser is a desktop application for turning point-cloud scans of historic vaults into projections, segmentations, geometric reconstructions, and measurement outputs. This site is the user guide for app users working through that workflow.

Project context

Vault Analyser forms part of the wider project Virtual Vaults: Using AI to process, analyse and supplement digital data, and contributes to the broader Tracing the Past initiative on medieval vault geometry and plan interpretation.

UKRI project record: UKRI955

UK Research and Innovation logo University of Liverpool logo Virtual Engineering Centre logo

What you can do in the app

  • Load an E57 scan and inspect it in 3D
  • Generate 2D projections from multiple viewpoints
  • Create segmentation masks for ribs, bosses, and other features
  • Reconstruct a 2D bay plan from the segmented vault
  • Reproject the interpreted geometry back into 3D
  • Extract rib traces and calculate measurements
  • Review a final geometric analysis summary
  1. Installation
  2. Before You Start
  3. Step 1 to Step 8

If you only need the software workflow, stay inside the Workflow section. The System page is only for readers who want a light technical orientation.

Nonstandard methods used in Vault Analyser

  • Gaussian projection rendering: point clouds are projected into smoother 2D images using Gaussian splatting rather than simple pixel binning, which helps preserve rib visibility in sparse scans. Related reference: Westover, 1990.
  • Prompt-driven segmentation: the segmentation stage uses SAM 3, so users guide the model with polygons, boxes, or text prompts instead of tracing every boundary by hand. Related reference: Ravi et al., 2024 for the Segment Anything family.
  • Template-based bay interpretation: Step 4 compares detected boss locations against starcut and circlecut-style geometric templates to support historical plan interpretation. Related background: Tracing the Past material on medieval vault plans.