Crowd detection AI that flags groups before they become a problem
A crowd rarely appears all at once. It starts as a few people, then a few more. Eye AI group detection runs on your existing CCTV cameras and notices when four or more people gather in one area. It logs the group, its size and how long it lasts, so your team can step in early and review busy periods later.
ILLUSTRATION WITH SAMPLE DATA
The problem
A gathering at a school gate, entrance or lobby is easy to miss when staff watch many screens.
After an incident, nobody can say how many people were there or for how long.
Peak times are guessed from memory instead of being measured.
How it works
- Choose the areas.In the free camera-readiness check we agree which camera views matter: a gate, a lobby, a forecourt, a corridor.
- Agree what counts as a group.By default Eye AI looks for four or more people close together in one area.
- The on-site computer watches.Analysis happens on a computer at your site. Your video stays on your network.
- You see the cluster.The event shows on your password-protected dashboard with group size, duration, time and a snapshot.
What you get
- Group events with the number of people and how long the group stayed together.
- A peak-time view showing when gatherings happen most.
- A searchable event history by day, hour and camera.
- CSV export for your own reports.
- A motion safety net that shows movement the system could not name, so nothing at your line is silent.
Where it is used
Honest limits
- This analytic is in beta. It has been shown working and tested on simulated movement, and real-footage validation is still under way, so we publish no accuracy figure.
- It flags clusters. It is not a precise head count of a dense crowd, and it is not a queue-length or wait-time measure. Queue wait times are on our roadmap and are not available.
- When people overlap heavily, some may be merged or missed, so the reported size can be lower than the real one.
- Distance, camera angle, poor light and glare all reduce reliability. We tell you during the camera check whether your view is realistic.
- A group is not a problem by itself. Eye AI tells you that people gathered, not why.
Questions about group and crowd clusters
Can it run on the cameras I already have?
Yes, if they provide a standard video stream on your network. The free camera-readiness check confirms it.
How many people make a group?
By default, four or more people close together in one area.
Does it count how long a queue is?
No. It flags a cluster and its duration. It is not a queue-length or wait-time tool.
Will it send me a phone alert?
Alerts to WhatsApp or email are a separate add-on that is launching, set up per site. Time-of-day schedules are not available yet.
Does the video leave my site?
No. Video is analysed on a computer at your site.
Tell us about your site.
We read every request and reply within 24–48 hours (working days, Monday to Friday) with questions and a proposed solution.
Request a solution