KINISI

Most of a Collection Has No Image Attached to It

In natural history, archaeological, and archival collections, most material has no image against its database record, where a database record exists at all. I've worked in collections; that is the ordinary state of things.

An object without a digital record is accessible only to researchers who already know exactly where it is. The expertise to document that material by hand is retiring faster than it's being replaced. The backlog is not getting smaller.

Moving fragile or valuable material off-site to get it imaged costs money before a single photo is taken: transport, specialist handling, insurance for the transit itself. The rig travels to the collection instead, so none of that applies.

A Robot That Photographs the Same Way Every Time

KINISI runs a fixed, preprogrammed sequence: a 5-axis robot holding the same angles, the same lighting, and the same order for every object on the rig. At volume, that consistency is the whole point.

Photography

Multi-angle stacks, merged into focus-stacked stills with full depth of field.

Video

The same sequence run as continuous footage, frame-identical to the stills pass.

3D Models

Photogrammetry or Gaussian Splatting reconstruction from the full angle set.

Swipe for photo, video, 3D →

Where This Was Tested

Swipe for more →

Every carabid type specimen Senckenberg held, sub-millimetre to over 4cm, captured on the same rig without re-tooling, delivered into Senckenberg's AQUILA database in months rather than the years a manual equivalent would have taken. I worked on that project. The collection happened to be beetles; the job (consistent capture, at volume, integrated into a working database) is identical whether the backlog is fossils, artefacts, or anything else worth keeping a record of.

One Pass Through the Rig, Four Different Uses

• 01Taxonomic reference imagery: the type-specimen image other researchers cite instead of requesting the beetle itself on loan.
• 02The AQUILA record: every image tied to a database entry, queryable by anyone with access.
• 03A condition baseline: the same captures document the specimen's state at the time, useful if it's revisited later.
• 04Publication-ready figures: usable in taxonomic literature, captured once rather than re-shot per paper.

Six More Objects, Same Rig.

Ten Objects an Hour.

That rate holds across a working run. Over a month it compounds to this:

Illustrative, based on a sustained rate of 10 objects/hour across an 8-hour day, 5-day week. Real collections vary: object size, fragility, and the formats you need all change the maths.

Objects Get Re-examined

• 01Collections get revisited (a new dating technique, a provenance question, a re-attribution, an exhibition nobody had planned when the object was first catalogued), so the capture has to be good enough to support questions nobody's asked yet.
• 02One master capture, done properly, can feed an archive record, an exhibition label, an AR model, and a published figure without ever re-touching the object.
• 03Delivered as archival masters (TIFF, AVI, FBX) plus ready-to-use compressed copies (JPG, MP4, GLB), benchmarked against FADGI or another named standard if your institution requires it. Full specification in the appendix slides ahead.

Four Steps From Shelf to Database

The same four steps apply on display and off it. Most of a collection is never exhibited, and once the rig is on-site there's no reason to stop at the pieces on the wall.

Numbers for Your Imaging Officer

• 01Photography: 24-megapixel captures, multi-angle stacks merged into focus-stacked stills, lit with cool-running LED, no UV (safe for fragile pigments, papyrus, textiles). A colour reference target is shot once per session; scale is overlaid from the rig's own coordinates rather than a ruler in shot. Larger or flat-format objects (paintings, murals, oversized panels) get multiple passes stitched into a single gigapixel image instead of one capped frame.
• 023D: texture resolution held at 4K. Mesh density is set per object; a coin and a cup don't need the same triangle count.
• 03Standards: benchmarked against FADGI, Metamorfoze, or your own imaging officer's standard, agreed before the job starts.

The Questions Your Registrar Will Ask

• 01Handling: done either by me directly or by the institution's own staff, whichever each institution's protocol requires. Every project to date has run with me working alone; no larger team stands behind the rig yet. Training is available if your own staff want to run the rig themselves going forward.
• 02Files: masters delivered as TIFF (photo), AVI (video), and FBX (3D), chosen to stay legible in decades rather than years, alongside a JPG/MP4/GLB compressed set for everyday use.

Nothing Stays With Me Afterwards

• 01Ownership: the museum owns everything delivered, raw captures included.
• 02Database: I don't build to one fixed schema. Senckenberg's integration worked because I adapted to AQUILA's own structure, not the other way round. Direct working experience with Darwin Core for natural history collections; the same schema-agnostic approach applies to CIDOC-CRM or whatever an archaeological collection already runs.

Text You Can Lift Into Your Application

Everything above argues this is worth doing. A funding panel needs a different case, in a different voice. Copy the paragraph below straight into your own application if it helps.