REAL

Holds up: a YOLO people counter with in/out and live totals is real, free code, though the demo's 'in' count lags

Facebook · Oct 2026

The claim'This AI model can count people entering and exiting and also show total people.' Overlay: 'Realtime In/out counter + unique people detection, built only with Python and OpenCV, DeepSort and YOLO.' 'Next Level Security!!' Offer: message for computer-vision services; free code on GitHub.

This one is what it says it is. The reel shows overhead CCTV footage of a busy jewelry store with boxes around each person, a live 'People' total, and in and exit counters, with the Python code scrolling underneath. That is standard, well-proven computer vision: an object detector (YOLO) finds people in each frame, a tracker gives each one an ID across frames, and a virtual line counts crossings. The code is public. The linked GitHub repo holds 19 small computer-vision projects under an MIT license, with about 225 stars, including a 'people-counter-in-out' footfall counter with a live dashboard. One detail is off: the overlay says DeepSort, while the repo lists ByteTrack as its tracker. Both are mainstream trackers, so it does not change the verdict. Two honest caveats. The demo's 'in' counter stays at 0 for the whole clip while people visibly move around, so the line placement or the crossing logic is not tuned for this camera; accuracy in a real store depends on camera angle, crowding and setup. And 'Next Level Security' oversells it: this counts bodies, it does not detect theft. What's being sold is the developer's freelance services, disclosed plainly in the caption. For a small shop, the free repo or an off-the-shelf people counter will get you footfall numbers.

What holds up

Reel transcribed and read frame by frame: overhead store CCTV with person boxes, a 'People' total between 11 and 16, 'in: 0' throughout and 'exit' climbing from 0 to 3, code scrolling below.
GitHub repo Nawaf-Rayhan585/YOLO_Projects: 19 projects, MIT license, about 225 stars and 97 forks, including people-counter-in-out (IN/OUT footfall counter with live dashboard), crowd-heatmap and queue-wait-time-estimator.
The repo names ByteTrack (via the supervision library) as its tracker; the reel overlay says DeepSort.
Caption discloses the offer: custom computer vision, AI/ML and websites by message, with an email and portfolio link.

What doesn’t

'Next Level Security' framing on a footfall counter that does not detect theft.
Demo's 'in' counter never moves, so accuracy on this footage is not shown.
Tracker named in the overlay does not match the repo.

The catch

Real, free, open-source people counting. Expect to tune the counting line and camera angle before the numbers are trustworthy, and don't buy it as a security system.

How to actually do it

Download the people-counter-in-out project from the MIT-licensed repo and run it on 10 minutes of your own door camera footage.
Count the same 10 minutes by hand and compare; move the counting line until the two match within a few people.
If you'd rather not run code, compare that result against an off-the-shelf door counter before paying anyone for custom work.

People counting with YOLO plus a tracker is mature, everyday tech, and this developer published working code for free. The only stretch is calling it security, and the demo itself shows the in-count needs tuning.

Confidence: Medium
Posted by: a freelance computer-vision developer page; reel sent to Buddy Sun Oct 11, 2026, ~7:46 AM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.