Bus & taxi operators

AI Passenger Counting on a Belize Bus: What the Camera Sees

A cabin camera watches the door and counts who gets on and off. Here is exactly what it catches, what it misses, and why the conductor's count still matters.

Oct 13, 2026 · 5 min read
Interior of a Belize bus showing the cabin door area where a ceiling-mounted camera counts passengers boarding and leaving

The 10 PM report nobody used to get

Six o'clock, the bus pulls in from the last Belize City run, the conductor hands over the cash, and that's the count. Whatever the fare box says, whatever the conductor remembers, that's the number an owner has had to work with for years. No second opinion, no way to check it against anything, no record once the night is over.

With a dashcam in the cabin running AI passenger counting, something new shows up in your inbox at 10 PM: a report. Boardings, alightings, an estimated count per trip, for every run the bus made that day. It doesn't replace the conductor's count. It gives you a second number to put next to it.

That's the whole idea. Not surveillance for its own sake — a second count, automatic, every single day, that didn't exist before.

How the camera actually counts: two lines, not one

The cabin camera isn't watching the whole bus and guessing how many heads are inside. It's watching the door, and it draws two lines across that doorway — one just inside, one just outside, a short corridor between them.

When someone crosses the outer line and then the inner line, that's a boarding. When someone crosses the inner line and then the outer line, that's an alighting. The software tracks each person through that corridor and, over the course of a trip, tries to match an exit to the entry it saw earlier — so it can estimate not just how many people got on, but roughly how far they rode, which is what the fare should be.

It's a simple idea built to survive a messy door. Two lines, one corridor, every crossing counted and matched. That's it.

What it catches well

On a normal stop — door opens, two or three people step up one at a time, the door line is clear — the two-line method does exactly what it's built for. Boarding counted, alighting counted, entry matched to exit, trip estimated. This is most of the stops on most routes, most of the day.

It holds up across a full route, not just a quiet stop. San Ignacio to Santa Elena, Belmopan to Belize City, the market-day runs along the Northern Highway with bags and boxes going up the steps — as long as people are crossing that doorwell more or less one at a time, the camera keeps up.

And it keeps up all day without getting tired, without losing track after the fourth stop in a row, without needing to glance down at a fare book while watching the door at the same time. That's the advantage over a count done by eye and memory — not that it's flawless, but that it's consistent, trip after trip, and it's written down.

Where it struggles: crowded doors, the doorwell, night runs

A loaded door is a hard door to count. When six people are getting off and four are trying to get on at the same time — Saturday market in Orange Walk, school letting out, a terminal rush — bodies overlap in that corridor and the camera can lose the clean line-cross it needs to tell one person from the next.

A conductor standing in the doorwell is its own problem. If the conductor's body is parked between the two lines for the whole stop, collecting fares and waving people through, the camera is trying to count people walking past or around a person who never actually crosses either line cleanly. The corridor gets crowded before the passengers even show up.

Night runs add a third difficulty. The cabin camera needs enough light to tell people apart at the door, and a dim stop on a dark stretch of highway gives it less to work with than a bus idling under terminal lights at noon.

None of this is hidden in the report. It's just the honest limit of a camera watching a door from one angle, on a moving bus, in the real conditions a Belize route actually runs in.

Why the conductor still matters

This is the part worth saying plainly: the camera does not replace the conductor, and it was never meant to.

The conductor knows the regular who rides free to church on Sunday, the two seats taken up by a mother and the infant on her lap who isn't a separate fare, the argument at the back door that the camera's doorway lines never saw because it happened at a window seat, not at the door. The conductor also collects the actual cash, which the camera doesn't touch and never will.

What the camera gives you is something the conductor can't produce on their own — a trip-by-trip, written, automatic second count that exists whether or not it was a good day or a rushed one. It's a check, not a replacement. The value isn't in the camera being right every time. It's in having two independent counts sitting next to each other, so a gap between them is something you can actually see, instead of something you can only suspect.

Reading the report without treating it as gospel

When the 10 PM report lands, read it as what it is: an estimate built from a doorway camera doing its best on real stops, real crowds, and real light conditions. It will track closely with the conductor's count most days. On the days it doesn't — a packed Friday run, a night trip on a dim stretch of road — that gap is information too. It tells you to look at that specific trip, not to throw out the whole system.

Used that way, over weeks and months, the report stops being a single number you either trust or don't, and starts being a pattern you can watch. That's the actual use of it — not catching one bad day, but noticing, over time, whether the two counts are tracking together or drifting apart.

Chat on WhatsApp