How accurate is Age recognition?
Age recognition is 85–90% accurate across seven age bands, measured against ground-truth audits at deployment.
Age recognition is 85–90% accurate across seven age bands, measured against ground-truth audits at deployment.
How the bands are structured
Age estimates fall into one of seven categories:
- Under 16
- 16–24
- 25–34
- 35–44
- 45–54
- 55–64
- 65+
Bands are wider than precise birthdays for a reason: estimating exact age from a single video frame isn’t possible, but classifying into a 10-year band is reliable, useful, and privacy-friendly.
How accuracy is measured
For each store, the setup team samples anonymous face-blurred snapshots from the cameras during the first week of operation and manually labels each one with a human-estimated age band. The AI’s estimate is compared to the human label and the resulting accuracy figure is what your dashboard reflects.
When accuracy is lower than the average
Stores with steep camera angles, very low light, or short-dwell entrance views can see slightly lower numbers. In those cases the setup team will reposition the camera view or switch to an alternative angle during tuning.