In active use Personal product 2026–present

Multi-Species Pelvis Morph Explorer

A browser-based scientific visualisation built around true glTF morph targets: one human base mesh, four great-ape targets and 62 corresponding landmarks. Visitors can select any species, move smoothly from the current form or play the complete comparative sequence.

  • glTF / GLB
  • Three.js
  • WebGL
  • Generalised Procrustes analysis
  • Thin-plate splines
  • 62-landmark correspondence
Concept illustration — no private project data shown.
selectable species forms
5
corresponding landmarks
62
shared mesh topology
1

Interactive comparative morphology

One topology, five pelvic forms

Select a species to move the shared surface smoothly from its current form. The optional landmarks reveal the 62 anatomical correspondences driving the comparison.

Only neutral species labels and anonymised geometry are shown. No specimen, collection or source metadata is included.

Current form Human
Drag to orbit · scroll to zoom

WebGL viewer

Load the comparative pelvis

The high-detail scientific surface is 50.6 MB. It loads only when requested, so the rest of the portfolio stays fast.

Morph to species

Reading the publish-safe species manifest…

01 · Problem

What needed solving

Comparative shape differences are difficult to understand from separate static renders. Changing camera, scale or mesh topology between images can make presentation differences look biological and hide which anatomical regions actually move.

Turn a familiar morphometric operation into a direct visual explanation: preserve correspondence, hold topology constant and let the viewer inspect the transformation from any angle.

02 · Constraints

The difficult parts

  • Preserve exactly the same 368,765-vertex topology across five substantially different pelvic forms.
  • Keep all 62 landmark correspondences synchronised with the surface during arbitrary transitions.
  • Make a high-detail 50.6 MB research asset optional so it does not slow the rest of the portfolio.
  • Provide orbit, zoom, direct species selection and sequence playback without obscuring the scientific comparison.
  • Publish only neutral geometry and labels while retaining private source and QC material locally.

03 · Approach

How the work was structured

  1. Used a joint alignment to place all species in a shared comparative coordinate system.
  2. Kept the human surface as the fixed base topology and encoded the other four forms as named GLB morph targets.
  3. Read target order, base species and sequence timing from a publish-safe manifest instead of duplicating those assumptions in the interface.
  4. Interpolated both GLB target weights and the corresponding landmark coordinates from the viewer’s current state.
  5. Added anterior reset, orbit and zoom controls, an optional 62-landmark overlay and reduced-motion behaviour.
  6. Deferred the large GLB download until explicit user action and kept the technical QC report outside the public build.

04 · Validation

What can be claimed now

The GLB contains one mesh named PelvisMorph, one shared topology, four correctly ordered morph targets and a complete five-species sequence. Local QC confirmed 62 corresponding landmarks per species, finite coordinates, sub-0.000016 input-unit landmark warp error and no embedded source paths, collection labels or specimen identifiers.

The published explorer moves continuously between any two current and selected forms, supports a manifest-driven 12-second sequence, keeps the landmarks synchronised and leaves the 50.6 MB surface unloaded unless the visitor opts in.

05 · Learning

What the project clarified

  • Morph targets solve the visual-comparison problem only when topology and vertex order are genuinely shared.
  • Landmark correspondence adds explanatory value beyond a surface animation because the anatomical relationship remains inspectable.
  • Scientific detail and web performance can coexist when expensive assets are optional and isolated from the main page load.
  • A public visual can remain scientifically meaningful without exposing the records used to construct it.

Next

Planned follow-through

  • Prepare a reduced-density web export while preserving the existing morph targets and landmark correspondence.
  • Add optional landmark-region explanations after confirming terminology suitable for a public audience.
  • Evaluate mesh compression after establishing a visually equivalent reference build.