Visual Context
ImageScope
We extract what the image shows and how it was taken. Contextual image metadata, structured and reliable.
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.
- 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.
- 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.
Send the submission
Photographs and scans go up together; the sorted, graded record comes back when analysis completes.
Route on confidence
A file with unplaced images, or a group whose confidence is low, reaches a person before assessment starts.
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