Damage Analysis
DamageScope
Visual damage analysis, end to end. From segmentation to repair estimation.
The problem
The assessment is not the slow part. Reading the photographs is.
An assessor opens a claim, works through the photographs one at a time, and writes down which panels are damaged before any judgement about the claim begins.
None of it is judgement, yet no two assessors work the same way, so the same photographs produce a different parts list, and a different estimate, depending on who opened them.
In and out
Photographs in, a damage record and a repair estimate out.
- Claim photographs
- Ordinary phone photography of the vehicle exterior, several angles of the same car.
- Nothing else
- No vehicle details to check against, no marked-up images, no capture template.
- Damage, located
- One entry per damage, traceable to the polygon on the photograph it came from.
- The repair call
- Repair or replace for each damaged part, with the method, labor hours, forecast cost and risk.
- A written summary
- The same findings in prose, ready to sit on the claim file.
Work is submitted and collected later; a full assessment expects several photographs of the same vehicle.
Capabilities
Exterior damage segmentation
Detect and mask all visible damage areas with pixel-level precision.
Damage type & severity analysis
Classify damage (dent, scratch, etc.) and assign objective severity scores.
Decisioning & Forecasting
Generate repair vs. replace decisions, labor cost forecasts, and verbal summaries.
The hard part
The hard part is not finding a dent.
The other party's car
The insured car and the third party's share a frame. Damage that cannot be tied to one of them is left out, not added to the wrong car's list.
Damage across a panel gap
A scratch crosses from wing to door; a crack runs from bumper into headlight. Each damage is assigned by how its shape overlaps the part, and weak assignments are dropped.
Severity that survives argument
Ask for an opinion on how bad a dent is and the answer moves between runs. This type, this measured share of this panel, therefore this band.
Where a person decides
It prepares the decision. It does not make it.
When it could not look
The record separates no damage found from could not be assessed, and names the photographs it was unable to use.
It checks its own account
The written summary is reconciled against the visual evidence part by part, and disagreement is surfaced rather than smoothed away.
You set where automation stops
Every value carries its own confidence score. Where the line between automatic and reviewed sits is your decision, not ours.
Who it is for
Insurance
Motor claims teams that need the same assessment quality on every file, not on a sampled few.
Fleet and rental
Return-condition records where every scuff carries a part, a severity band and the hours to put it right.
Inspection and remarketing
Condition reports produced to one standard, with the evidence for each finding attached to its image.
How it fits
Start with one workflow, not the whole platform.
Send the photographs
Submit a claim's images through the API and collect the record when the assessment is finished.
Route on confidence
Put the threshold where your risk appetite sits: clear files proceed, borderline files reach a person with the reason attached.
Keep the record
The assessment is structured data, exportable as a document, and it records which classification scheme produced it.
Photographs can be added to an existing file and the assessment re-run, so a claim arriving in stages stays one record.
Questions we get asked
What is DamageScope?
DamageScope is Archmir's vehicle damage assessment workflow. It takes photographs of a damaged vehicle and returns a structured record: each damage segmented at pixel level, attributed to a named exterior part and graded for severity, with a repair-or-replace decision, labor hours, forecast cost and risk.
What do I need to send?
Ordinary photographs of the vehicle exterior, taken on a phone, with several angles of the same car. No capture protocol, no marked-up images and no vehicle details are required.
Does it replace the assessor?
No. It removes the mechanical pass, so an assessor opens a prepared record instead of a folder of photographs. Where files clear automatically, that threshold is yours to set.
How do I know it is right?
Severity is derived from the damage type and the measured proportion of the panel affected, so the same photograph produces the same band every time. We publish no accuracy figure; measure it on files you have already settled.
What does it not do?
It assesses the visible exterior. Damage beneath a panel is flagged for physical inspection rather than confirmed, and photographs too poor to read are reported, not guessed at.
Where to start
Take a folder of claims you have already handled and compare the record, the hours and the estimate against what your assessor wrote.
Initialize DamageScope