Evaluate AR inspection with one real part and explicit acceptance criteria

1. Fix the inputs and baseline

Choose one stationary rigid part and provide STEP / IGES, dimensions, surface information, working distances and lighting conditions.

Fix checkpoints, defect samples and the manual inspection baseline. Agree how to measure missed issues, false prompts and completion time under comparable conditions.

2. Validate tracking and checks

Record initial alignment, usable overlays, walk-around motion, occlusion, tracking loss and recovery. Test repetitive structures, wrong parts and wrong sizes separately.

Remove or misplace inspection accessories to check whether the inspection workflow identifies the issue and whether reference alignment is affected.

Record inspector decisions separately from system prompts. Use appropriate gauges for dimensional reference measurements.

3. Complete the reporting workflow

Import, author, publish, download, inspect, generate a report and review it in the browser using the same project revision.

Run an offline inspection and check session, photo and report completeness after reconnecting. Confirm permissions, storage and diagnostic collection scope.

4. Agree the expansion criteria

Report completion time, missed issues, false prompts, unavailable periods and evidence completeness by part and scenario. Retain failure examples.

CADAR is an integration prototype. Engineering targets for initialisation and drift must become site acceptance items, not claims of achieved precision.

Decide the next part numbers and training plan from pilot evidence and mutually accepted results.