AI People Counting Camera and Software for Manufacturing Sites
acty.dev builds AI people counting systems that turn camera footage into live occupancy, entry/exit, and zone counts for manufacturing and warehouse sites, validated in a pilot on the site's own footage before it is rolled out.
AI people counting software turns a camera watching a work zone into a live count: it knows the moment a posted occupancy limit is crossed, without anyone doing a headcount. The same footage builds a timeline of which stations are staffed and which stand empty across a shift, flags a post that’s sitting empty when it shouldn’t be, and turns foot traffic between zones into a picture of where a floor gets congested. The site gets counting tuned to its own zones and shift patterns during a pilot, not a one-size-fits-all occupancy tool.
Where it applies on the floor
-
Occupancy limit in a work zone. A work zone often carries a posted occupancy limit, set for safety or simply because the space can’t hold more people without slowing everyone down. The system counts how many people are inside that zone continuously, and as soon as the count goes past the posted limit, it raises an alert rather than waiting for someone to notice the crowd. That turns a limit written on a sign into something acted on in real time, not a rule people quietly work around during a busy shift.
-
Presence timeline at a station or machine. Across a shift, staff move between a station or machine, other work zones, and rest areas, and today that pattern is invisible unless someone happens to be watching. The system builds a timeline of presence and absence tied to each zone, so a stretch away from a post that runs longer than the normal pattern for that role stands out on its own. A manager reading that timeline sees staffing balance and can compare one shift against another without walking the floor to check.
-
An unattended post or a zone with no assigned task. Some posts are meant to be staffed at all times, and some zones are meant to be empty unless someone has a specific reason to be there; either state slipping into the other is usually a coverage problem or something worth a second look. When someone lingers in a zone without an apparent task, or a post that should be staffed sits empty, the system flags the anomaly and sends a notification with a clip of the moment. That gets attention to a coverage gap or an out-of-place stay while it’s still relevant, instead of surfacing in an end-of-shift report nobody reads closely.
-
Zone load for layout decisions. Over the course of a shift, people trace real paths between zones, entrances, and workstations, and those paths rarely match the layout a floor plan assumed when it was drawn. The system aggregates zone and line-crossing counts across the area it watches, so a congestion point at one crossing, or a zone that sits underused despite being staffed for it, shows up as a pattern rather than an impression. Layout, staffing, and routing decisions then start from what actually happens on the floor across a shift, not from where the floor plan expected people to go.
The problem we solve
Knowing how many people are on a floor, in a zone, or moving through an entry point usually depends on manual headcounts, badge logs, or guesswork, which makes occupancy limits, congestion, and zone capacity hard to track as conditions change during a shift.
How we work
- Pilot scoping.The zones, entry points, and occupancy limits this site needs counted get agreed on before any build starts; so does the control sample used to check those counts against them.
- Pilot on your footage.Counting then runs against your own footage, tuned to the criteria the pilot process on the AI video analytics page sets for every area.
- Integration.Confirmed entry, exit, and station-time data feeds straight into the systems your team already uses for shift planning, with an alert to whoever runs the shift, by messenger or dashboard, when a zone goes over its limit — all confirmed before rollout.
- Rollout and support.Documentation and a full handoff go to your team as counting extends to the rest of the zones and entry points agreed in scope, once the pilot is confirmed, with support continuing afterward as shift patterns or zone layouts change.
What you bring / what you get
Inputs
- Existing CCTV access if you have it, otherwise footage clips covering the zones and entry points you want counted
- The zones, entry points, or occupancy limits you need counted
- Which dashboards or messenger channels should receive your counts and alerts
Outputs
- Zone occupancy and entry-point counts confirmed on your own footage against the agreed limits
- Live occupancy, entry/exit, or zone counts connected to your dashboard or alerting systems
- Handoff documentation your team can act on, listing every zone, entry point, and count path in scope
Definition of done.A people counting pilot checked on this site's own zones, entry points, and occupancy limits against a control sample and the acceptance thresholds agreed for the pilot; confirmed counts feed your dashboard or occupancy-alert systems, a presence timeline exists per zone, and the rollout is documented.
Fit and anti-fit
Good fit
- Sites tracking zone occupancy, station time, or shift patterns whose cameras already see the zones to count
- Operations leads who want counting validated against real traffic patterns before rolling it out further
Not a fit
- Zones or entry points that no camera can see today, with nothing planned to change that
- Locking in a counting-accuracy figure before footage has even been reviewed
Questions
- What does an AI people counting system include?
- Occupancy alerts against a posted zone limit, a presence timeline showing time at a station and time away from it, a flag when a post sits unattended or someone lingers without an apparent task, and a zone-load picture built from entry, exit, and crossing counts, each proven on this site's own footage first, then wired into the dashboards and alert channels your team already watches.
- Can it flag when a zone goes over its occupancy limit in real time?
- Yes. The moment the number of people in a monitored zone crosses the limit posted for it, an alert goes out on its own, so the shift lead hears about the crowd while it is still forming.
- Can you tell how long someone spends at a station or in a zone?
- Yes. The system builds a presence timeline for each zone or station, showing time at post and time away from it, which makes it possible to compare shifts and roles side by side instead of relying on a supervisor's impression.
- Can you guarantee a specific counting accuracy?
- No. Accuracy depends on your camera placement, lighting, and the zones you need counted, so acceptance thresholds are agreed and tested during the pilot rather than promised in advance.
- How is this priced?
- Which zones, entry points, and limits get counted is scoped in an initial task review, and the pilot itself has to clear the thresholds set in that review. Once it does, quoting wider site coverage and the tuning that keeps it accurate across shifts becomes a separate, later step.
Explore
Have a process, dataset, or camera footage to work from?