VISION | P AI
Measuring audience attention, with privacy at the core. Turn your existing cameras into objective data on footfall, dwell times and audience composition.
How long do people really stop in front of that artwork?
VISION | P AI turns the footage from your existing security cameras into precise measurements of how the audience experiences a space. No new hardware, no image storage, no compromise on privacy.
Only anonymous data — the kind that helps whoever designs and runs that space: museums, retail, public and cultural venues.
Four metrics, one goal: real attention
Objective numbers instead of gut feeling. Every metric is computed on anonymous data.
Footfall
Precise count of every person crossing the frame, whether they stop or not. Real flow by hour and by day.
Dwell time
Time spent in front of an artwork or display, measured to tenths of a second with precise start/end timestamps.
Audience composition
Aggregated demographic estimates (men, women, children) as statistical data. Never individual identification.
Heat maps
Visual overlay showing where the audience concentrates within the space, on a floor plan or directly on the actual frame.
Five steps, zero fuss
From an existing recording to data in your browser, in minutes.
Existing footage
Upload the recordings from your already-installed security cameras.
AI detection
Anonymous tracking of people frame by frame, with temporary identifiers.
Zones of interest
Define the areas in front of artworks or windows via a reference image.
Dwell tracking
Stops shorter than 3 seconds are not counted as attention.
Data at hand
Query the anonymous results from any browser, no software to install.
Privacy isn't a footnote: it's the architecture
VISION | P AI is built from the ground up not to handle personal data. No faces, no archives, no identification. Only anonymous numbers describing a collective phenomenon — the attention of the audience.
- No facial recognition, no face stored
- No image or video archive on the platform
- No individual identification
- Only anonymous events: times and durations
- Temporary identifiers deleted after processing
- Analysis runs on-site — the platform only receives numbers
Three worlds, one tool
Museums and cultural institutions
- Identify most and least engaging artworks
- Map flows across rooms and paths
- Verifiable data for grants and sponsorships
Retail and commercial spaces
- Dwell times in front of shop windows
- Peak hours to plan staffing
- Compare before and after refits or renovations
Public spaces and services
- Hourly footfall by day of the week
- Congestion and waiting zones
- Data-driven staffing decisions
Available everywhere, with nothing to install
Browser access
Reserved access via personal accounts, device-independent, zero installation.
Cross-filter search
Combinable filters by zone, demographic segment, minimum dwell time and time period.
Multi-camera, multi-site
Each additional space is configured as a new zone. Client data always kept isolated.
Numbers from actual runs
Sample processing on museum corridor recordings, to give a sense of scale.
| Scenario | Passages | Dwell events | Avg duration |
|---|---|---|---|
| Corridor · 8h 20m | 295 | 111 | 20.5 s |
| Corridor · 6h 30m | 228 | 100 | 18.9 s |
| Corridor · 1h | 29 | 26 | 18.6 s |
What VISION | P AI does NOT do
Because it matters to know this up front.
Trend analysis: not an exact head-by-head census
Demographic estimates are reliable as an aggregate, not for the individual
Camera position affects accuracy (high angles or long distances reduce precision)
In rare cases loss of tracking can cause a double count when the person reappears
A one-week pilot on your own footage
We analyze a sample of your recordings and return real engagement data: passages, dwell times, peak hours, attention maps.