Stage 0 · @dpa/capture
Creating record, not querying it
Every other stage in this system begins after an object has already been recorded by somebody. Where no record was ever created, all they can do is measure the hole. This is the stage that fills it.
What this page is
A working demonstration of the rubric and the real-time guidance loop, driven by sliders rather than a camera. There is no mobile capture client yet and nothing here measures a photograph. What is real is the logic: how ten measurements become one fitness class, one instruction, and a record that states its own limits. Every number on this page is computed live by @dpa/capture.
Why this is the priority
The Bura askos scores 58 with 0 of 9 registers able to name it. That is not a gap in our searching. Bura funerary sites were never inventoried, and INTERPOL, the FBI National Stolen Art File and the Carabinieri TPC archive hold reports of thefts from documented collections. An object from an unrecorded site cannot appear in them, ever, however much is later learned about it.
The object’s own timeline records the trap exactly: Niger ratified the 1970 UNESCO Convention in 1997, but the Convention “requires a pre-existing inventory — which Bura sites lack.”
Capture is the only mechanism in this programme that creates record rather than querying it. It cannot help an object already looted from an unrecorded site. It can start the record, from today, for everything still in museum, community or private custody — which is most of the material this programme exists for.
Three numbers, three questions
| Number | Asks | Owned by |
|---|---|---|
confidenceScore | How much is known about where this object came from? | @dpa/assess |
coverage | Could the registers have known anything at all? | @dpa/assess |
qualityScore | How good is the record we just made of the object itself? | @dpa/capture |
These never mix. @dpa/capture may not import @dpa/assess and @dpa/assess may not import @dpa/capture; a test asserts it at the module boundary. Otherwise a museum could raise an object’s provenance confidence by buying a better camera, and an institution with a well-funded imaging department and no documentation would outscore a community holding thorough written records and a phone.
Start from
A museum registrar with no photogrammetry training, following the on-screen instruction one step at a time.
Aiming for
If you stopped now
indicative
This session would currently be rated indicative; you are aiming for study. 1 further check runs after processing and cannot be judged here.
Do this next
Move closer, or use a longer lens. Fill more of the frame with the object — the resolution of the record is set here and cannot be recovered later.
Limiting dimension: Ground sample distance
Overall capture score — 68 / 100
A progress meter, not a fitness rating, and not a provenance score. The class above is what governs use; this number exists so a contributor re-shooting a session can see movement before the class flips. The two are allowed to disagree — set a reference capture and then remove the scale bar to see it happen.
The ten dimensions
Nine can be measured while capturing. The tenth cannot be known until the photographs are processed, and is shown separately for that reason — guidance that waits for it is a post-mortem, not advice.
88% — permits study
Surfaces you never pointed the camera at are reconstructed by guesswork or not at all. For a funerary vessel the underside and interior lip are exactly where potters' marks, repair seams and excavation damage sit — the features that distinguish this object from another of the same type.
88% — permits study
Distinct from angular coverage: you can orbit an object fully and still never see inside a cavity or behind a handle. Occluded regions become holes in the mesh, and a hole where a maker's mark should be is indistinguishable from an object that never had one.
72% — permits study
Photogrammetry works by matching the same physical point across several photographs. Below roughly 60% overlap the matcher loses the thread and the reconstruction splits into fragments that cannot be aligned. CIPA guidance for close-range heritage work is 60–80%.
VoL 260 — permits reference
Motion blur is the single most common reason a phone capture fails, and it is invisible on a small screen at capture time. Blurred frames do not merely add noise — they contribute wrong feature matches, which bends the reconstructed geometry rather than softening it.
1.2% clipped — permits study
Clipped highlights and crushed shadows contain no recoverable detail. On glazed or burnished ceramic a blown highlight erases precisely the surface it sits on, and no amount of processing brings it back — the information was never recorded.
0.55 mm/px — permits indicative
The real resolution of the record: how much of the object one pixel covers. It sets the smallest feature that can ever be recovered. Tool marks, incision and slip decoration on West African terracotta are sub-millimetre; a 2 mm/px capture cannot see them and never will.
AR-derived scale — permits study
Without a calibrated scale in shot, the model has shape but no size, and an object with no size cannot be measured, compared against a written description, or matched to an excavation record. Phone AR tracking gives approximate metric scale; for anything intended as a reference record a physical scale bar remains the standard.
none — permits study
Condition assessment depends on colour being comparable across captures years apart. Without a colour target in shot, the record carries whatever cast the room lighting had, and a later capture cannot distinguish genuine deterioration from a different set of light bulbs.
78% — permits study
If lighting changes while you orbit, shadows move across the surface and the matcher treats a moving shadow edge as a physical feature. This bakes shading into the texture and can emboss shadow edges into the geometry itself.
92% — permits study
Focal length, sensor size and camera pose let a reconstruction be checked, and later re-run with better software. Frames stripped of metadata — as most images shared through messaging apps are — can still be used, but the result can no longer be independently verified.
What the resulting record says about itself
Generated by buildCaptureRecord from the numbers above, for a capture of the Bura askos — a funerary vessel — held by the museum that holds the object.
Attests
An object was observed at Musée National Boubou Hama, Niamey on 2026-02-11 by an operator acting as custodian institution, whose relationship to the object was checked by institution attested. The record of that observation is rated indicative: Fit to display. Not fit to measure, compare or identify against. Shows what the object looks like; proves nothing about it. The dimension currently holding this back is ground sample distance. The resulting assets are bound to the exact set of source photographs they were reconstructed from, and every one of those photographs carried a verifiable capture-time seal.
Does not attest
This record does not attest that the object was lawfully excavated, exported, acquired, or is lawfully held. It records an observation of an object, not a right to it. Capture quality and provenance legitimacy are unrelated: a high-quality capture of an unlawfully held object is a high-quality capture of an unlawfully held object. This capture is rated indicative and is not sufficient to serve as the identifying record of the object: it cannot reliably distinguish this object from a similar one, nor detect substitution.
This field is mandatory in the schema and buildCaptureRecord throws if the legitimacy disclaimer is missing from it. A capture record carries a signature, a hash, a timestamp and a chain of custody, which is exactly what proof of lawful ownership looks like. It is not that.
Asset disclosure
- mesh → source-community
- Funerary material. Held at source-community tier until the community with cultural authority over this object has been consulted. Publication is a consent question, not a configuration setting.
Where the chain breaks today
C2PA solves capture-time provenance for photographs well — a hard binding over the asset bytes, an X.509 signature, an RFC 3161 timestamp, and in Truepic’s implementation a signature applied inside the secure enclave before the image leaves the sensor.
None of it survives photogrammetry. As of spec v2.1–v2.4 the normative format list is JPEG, PNG, GIF, TIFF, BMFF video and PDF. glTF/GLB, USDZ, E57 and PLY are absent. The photographs can be sealed to a very high standard and the mesh built from them inherits none of it. The chain breaks at exactly the step that produces the artefact anyone will actually look at.
Every crowd-sourced heritage capture effort we surveyed — Rekrei/Project Mosul, Backup Ukraine, the Million Image Database — used general-purpose photogrammetry apps and sealed the 3D output not at all. Rekrei accepts any photograph with no submission quality protocol whatsoever.
ReconstructionBinding is our answer: one signed structure binding the source image set, how many of those carried a verifiable seal, the pipeline and its parameters, the output mesh hash, and a perceptual hash as a soft binding so a re-exported copy can be re-associated. It does not make reconstruction reproducible — photogrammetry is not bit-deterministic — and chainComplete says so rather than implying a guarantee the format cannot make.
One citation we could not verify
The feedback that prompted this work cited a “Manhattan Bridge pedestrian capture protocol.” No protocol of that name exists in any public source we could reach — not the Library of Congress HAER record for the bridge, not Starling Lab, not C2PA, not NYC DOT. The closest real thing is Starling Lab and Numbers Protocol’s Starlingcapture, which anchors a cryptographic birth certificate in device hardware at capture time. We are not treating it as prior art until the original source is identified. The design above is grounded in what we could verify: Apple’s ObjectCaptureSession, the C2PA specification, and the London Charter.
What is not built
- No metric extractors. The rubric is defined and nothing computes it from actual photographs yet.
- No mobile client. Whether to build one, extend VANGO, or wrap Apple’s Object Capture is an open decision — and it reopens ADR-007, which placed VANGO outside the core on the premise that capture was out of scope.
- No issuer class for a contributor. Someone who scans an object they do not own is neither a pseudonymous holder nor an accredited institution. ADR-004 is reopened.
- Consent is unresolved and now blocking. Defaults are set closed — funerary and sacred material is held at source-community tier — which buys time and is not a substitute for asking.
- Thresholds are ours. There is no ratified cross-body numerical rubric for heritage 3D capture; the London Charter, the most widely adopted framework, is a principles document. Every band sits in one file so the argument can be had against specific numbers.