Big News: Archmir is now officially a Google Cloud Scale AI Tier Partner, accelerating our innovation with enterprise-grade AI infrastructure.Learn More

The problem

Somebody sorts the pile before anyone assesses anything.

A claim file arrives as a pile of phone pictures in no order: some of the insured car, some of the other party's, some of the accident report and the registration.

A handler separates them, works out which car is which, and judges whether the light and the reflections leave anything assessable. That judgement is never written down.

In and out

Every photograph placed, grouped and graded.

What goes in
The whole submission
Every photograph and scan from the claim, sent together rather than piecemeal.
Nothing else
No capture protocol, no shot list, no note about which car is which.
What comes out
A sorted file
Every image placed as vehicle imagery, paperwork or unplaced, and labelled by what it shows.
Vehicles, separated
Photographs grouped per physical vehicle, each group with a confidence score and a written reason.
Capture conditions
How each photograph was taken, recorded on the file instead of judged again by the next person.

The grading lands at intake, while the vehicle is still there and a retake still costs nothing.

Capabilities

Environment & weather classification

Detect the scene (indoor/outdoor) and weather conditions from visual cues.

Lighting & blur analysis

Analyze image conditions, reflectivity, and clarity for quality assessment.

Framing & proximity detection

Identify camera angle, distance, orientation, and visible vehicle regions.

The hard part

Grading one photograph is easy. Reading a set is not.

One car or three

Whether a photograph is sharp is a question about that photograph. Which of three cars it shows is a question about the whole submission, answered in one pass.

Grouping checked against pixels

A grouping is a proposal, not a result. Each photograph-to-vehicle association is checked against what was actually detected in that image, and dropped where it does not hold.

No image falls through

An image that fits no group is recorded as unplaced, not dropped. A photograph that quietly disappears returns a clean file that was never examined.

Where a person decides

The grading is evidence, not a decision.

The call stays yours

A photograph that cannot carry an assessment is named on the file; whether that means a retake or an exception is your rule.

It declares a shot unusable

With no vehicle in frame, or a cabin shot where an exterior view is needed, no damage assessment is attempted.

Cannot determine is an answer

Every axis carries an explicit unknown value, so the file records that a condition could not be read rather than guessing.

Who it is for

Insurance

Motor claims intake that arrives sorted and graded, so triage is a lookup rather than a first pass.

Fleet and rental

Handover packs audited rather than trusted, so an incomplete set is caught at the branch, not later.

Inspection and remarketing

Condition reviewers get the axes they argue over stated on the file: blur, reflection, and cabin against exterior.

How it fits

It runs at intake, before anyone opens the file.

  1. Send the submission

    Photographs and scans go up together; the sorted, graded record comes back when analysis completes.

  2. Route on confidence

    A file with unplaced images, or a group whose confidence is low, reaches a person before assessment starts.

  3. Keep the reasoning

    Each group carries the written reason those photographs were treated as one vehicle: the record a disputed settlement needs.

The sorted, grouped file is what DamageScope and ArchDoc work from, which makes this the natural place to start.

Questions we get asked

What is ImageScope?

ImageScope is Archmir's image triage workflow. It takes the photographs submitted with a claim and returns a sorted file: which images show a vehicle and which are paperwork, which photographs show the same car, and the scene, weather, lighting and camera position behind each shot.

What do I need to send?

The photographs and scans as they were submitted, in one go. No capture protocol, no shot list, and no note about which pictures belong to which vehicle.

Does it replace the claims handler?

No. It removes the sorting pass, so a handler opens a file already separated, grouped and graded. Where files clear automatically, that threshold is yours to set; the handler decides what the grading means.

How do I know it is right?

Every grouping carries a confidence score and a written reason, and is checked against what was actually found in the image before it is kept. We publish no accuracy figure; test it on claims you have already closed.

What does it not do?

It does not assess the car: no damage, no plate, no document fields, because those are other workflows. It grades and groups the photographs, then hands the file on; it settles nothing.

Where to start

Open a claim you have already closed and compare the sorted file against how your handler actually did it.

Initialize ImageScope

Other workflows