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

The dispute is about evidence, not damage

Handover damage disputes are rarely about whether a scratch exists. They are about when it appeared, and the record that would settle it is usually a handful of photographs taken in a dim car park with no consistent angle, lighting or coverage.

The failure is at capture time, and no amount of downstream analysis fixes a photograph that was never usable.

What changes at handover

Capture is validated on the spot. ImageScope reports lighting, blur, reflectivity, camera angle, proximity and which section of the vehicle is visible. An inadequate photograph is caught while the vehicle and the customer are both still present.

The vehicle is identified from the image itself. CarScope reads the plate and VIN and profiles the vehicle, so the condition record is bound to the actual vehicle rather than to whatever was typed into a form.

Condition is recorded structurally. DamageScope segments every visible damage area at pixel level and labels the exterior part it belongs to, across 30+ parts. The record is a structured object, not a photo album.

Why part-level masking matters here

A per-photograph comparison cannot answer "is this the same dent." A per-part comparison can. Because damage is masked at pixel level and attributed to a named exterior part, the check-in record can be diffed against the check-out record part by part — and the output of that diff is specific enough to put in front of a customer.

Operational effect

The gain is not primarily in detection accuracy. It is that the condition record becomes machine-readable, comparable and consistent across branches, drivers and lighting conditions — which is what turns a subjective argument into a lookup.

Solutions

Who takes the photographs?

Whoever is at the handover — a branch employee, a driver, or the customer on their own phone. ImageScope grades each capture on the spot, so a photograph that is too dark, too close or too blurred to support an assessment is rejected while the vehicle is still there.

How is a new scratch distinguished from an existing one?

By comparing segmentation output between the check-out and check-in records for the same vehicle. Because DamageScope masks damage at pixel level and identifies the part involved, the comparison is per-part rather than per-photograph.

Does this work across a mixed fleet?

Vehicle identification is model-driven rather than template-driven — CarScope profiles make, model, year, body type, colour and trim from an image. There is no per-model configuration to maintain as the fleet changes.

Built on

Initialize System Access