More cameras than people to watch them.
Cannabis operators install cameras because they have to and because the product is worth stealing. The result is a site where nearly every room, door and fence line is on video, and a monitoring screen packed with tiles that look almost identical from hour to hour: grow rooms under the same lights, a vault door that never moves, a dark parking lot, a fenced yard with a trash enclosure.
The guard or remote operator on that screen is not short on video. They are short on attention. Semaphor is AI video monitoring assistance that reads the same screen they already watch and tells them which tile needs a look right now.
What guards actually watch at a cannabis site.
Post orders differ between a storefront dispensary and a cultivation or processing building, but the camera positions that carry the real risk are consistent:
- Vault and product storage entries. The door, the corridor leading to it and the room itself. After close, nobody should be there.
- Sales floor and point of sale. Busy during business hours, empty and locked after. The after-hours version of this view is the one that matters.
- Delivery and loading doors. Transport pickups, deliveries and waste removal happen here, often on a schedule the guard knows.
- Perimeter fence and parking. Grows and processors often sit in industrial or rural areas with long fence lines and poor lighting.
- Roof, HVAC and utility access. Less obvious entry points that burglars test first.
During business hours the job is watching behavior. After close it becomes a presence question: is anyone at the vault door, the back fence or the roof hatch who shouldn’t be?
Events that matter. Activity that doesn’t.
A person at the vault corridor after the last employee badges out is an event. A vehicle idling behind the building at 3 a.m. with its lights off is an event. Someone walking the fence line and stopping at the same spot twice is an event.
A grow-room fan cycling on, a timer switching lights, an oscillating leaf in front of a lens, a cat crossing the lot, and the same employee car parked in the same spot all night are not. On a motion-only system those nuisance triggers bury the real ones, and after a few weeks operators stop trusting the alarms at all.
Camera and layout problems specific to cannabis.
- Lighting cycles. Grow rooms swing from intense light to dark on a schedule, and some rooms use unusual spectra. Pixel-change alarms fire every cycle.
- Dense layouts. Compliance-driven coverage means big grids of small tiles. A person in one small pane is easy to overlook.
- Cycling views. Many dashboards rotate through camera pages. A camera that cycles off-screen goes unwatched without anyone noticing.
- Long evidence trails. When something happens, owners, investigators and insurers want clean, original footage with a clear record of what the guard saw and when.
What your guards watch
- Grids of near-identical grow, storage and hallway views
- A vault door that should never move after close
- Dark fence lines and back lots at remote sites
- Lighting changes and fans that trigger motion alarms
What Semaphor surfaces
- A person in the vault corridor, flagged on the pane where it happened
- A vehicle stopped behind the building, not the parked employee car
- An AI scene report with a recommendation the guard can act on
- Explicit notice when a camera cycled off-screen and went unobserved
How Semaphor helps a cannabis monitoring post.
Semaphor works from the camera software you already use. It captures the monitoring screen, separates each camera pane and follows those cameras as the layout cycles. No camera credentials, no new VMS and no integration project with the system your compliance consultant specified.
Local person and vehicle detection watches every pane at once. A lighting cycle alone does not become an alert; a person or vehicle does. Duplicate suppression and cooldowns keep that one staff car from alerting all night.
When something real appears, an activity alert lands in the guard inbox immediately. A vision-language model then reviews the scene and updates the same card: what is happening, what it observed, whether guard review is suggested, and what remains uncertain. The guard decides whether to call the owner, dispatch patrol or contact police. The AI never does.
Every alert keeps the original frame beside the AI assessment, with an exact SHA-256 hash and exportable manifest. That evidence and audit trail is what you hand to an owner, investigator or insurer. For how to package it, see our guide to incident report video evidence.
No facial recognition. Semaphor tracks people as numbered, per-camera tracks, never as identities. That keeps privacy reviews simple at sites that already live under heavy scrutiny.
Built for the companies that guard these sites.
Most cannabis operators contract with a security guard company or a remote monitoring center rather than running their own control room. Semaphor is priced and built for exactly those operators: cover more client cameras from one station without adding seats. Pair it with our guide to after-hours trespassing to tighten overnight post orders.
The fastest way to see it is on your own screens. Start a pilot on one monitoring station and measure missed events, nuisance alerts and response time on your real client sites.
Cannabis camera monitoring FAQ
Do we have to replace our compliance-approved camera system?
No. Semaphor reads the rendered monitoring screen, so the cameras, recorder and VMS stay exactly as installed. Nothing about your existing recording setup changes.
Will grow-room lighting cycles flood the inbox?
Alerts are driven by people and vehicles, not raw pixel change. A lighting change with nobody in the room does not produce an actionable alert, and cooldowns stop repeat cards for the same object.
Can Semaphor lock doors or call police?
No, by design. It surfaces the event with evidence and a recommendation. Your guard makes the call and dispatches through your existing procedures.