Ground-truth audit
A ground-truth audit is the manual verification process used to confirm Aura Vision's accuracy: a human counts people in a sample of video clips and the result is compared against the AI's count for the same period.
Ground-truth audit is the manual verification process used to confirm Aura Vision’s accuracy. A human counts people in a sample of video clips from a store’s cameras, and the result is compared against the AI’s count for the same period. The accuracy figure reported for each store comes directly from these audits.
When audits happen
- At go-live. Every store is audited against ground truth during initial setup, before the data is signed off.
- Quarterly thereafter. A random sample of stores is re-audited every quarter to catch model drift from lighting changes, seasonal layouts, refits or new uniforms. Samples include both peak and off-peak hours.
What’s checked
Audits cover entry-count accuracy, the staff-versus-customer split and demographic estimation. When a discrepancy is found — usually an obstructed view, a sub-optimal angle or a repositioned camera — the fix is typically a small adjustment and accuracy is restored within days.