Back to Publications
PUB-019/Case Study/Sep 2026

Simam3D: six evidence-led case studies in practical spatial AI

A public index of six Simam3D studies covering evidence-first reconstruction, native Gaussian export, multi-view support, uncertainty-aware scene hypotheses, media-to-3D ingest and consent-first digital presence capture.

Business value

The series demonstrates a reusable approach to spatial AI: expose provenance, bound uncertainty, start with ordinary media, and make governance part of the workflow. These patterns are relevant to asset documentation, inspection, digital twins, virtual production and spatial data pipelines.

Evidence and measures
  • - Six distinct public-safe studies are separated by research question and maturity rather than presented as one undifferentiated demo.
  • - The hosted Simam3D record includes a verified 14-output evidence bundle and a local regression gate reported at 78 tests.
  • - A verified DA3 run reported 63,504 candidate Gaussians and exported 2,048 after deterministic capping; this is an export measure, not a quality claim.
Research Questions
  1. 01What evidence should accompany a spatial asset before it enters an operational workflow?
  2. 02How should uncertainty and viewpoint support change the next capture decision?
  3. 03Which consent and provenance controls belong in a digital-presence system from its first prototype?
Method
  1. 01Separate observed evidence, model inference, synthetic priors and generated hypotheses in both language and metadata.
  2. 02Record the input, model path, runtime, hashes, camera-pose source, confidence and fallback behaviour where available.
  3. 03Publish limitations and the next validation question alongside the visual result.
Evidence
  1. 01Public Simam3D GPU Space with hosted verification and downloadable evidence outputs.
  2. 02Six source case studies in the Simam3D case-study collection, covering the Evidence Lab through SimamTwin3D.
  3. 03Public project records linked below, each labelled Live, In Benchmark or Prototype.
Next Steps
  • Run named real-capture and held-out-view evaluations before making comparative quality claims.
  • Test the evidence labels with technical, creative and operational users.
  • Define a governed consent and retention review for future digital-presence experiments.
Related live work
Simam3D Evidence Lab ->Direct Gaussian Export with DA3 ->Multi-View Fusion Evidence ->What's Behind the Image? ->SimamAnything2GS ->SimamTwin3D ->Hosted Simam3D GPU Space ->
Related Case Study records