Restaurant camera analytics: put the cameras you already paid for to work

From evidence to operations

A security camera is a backwards-looking tool: something goes missing, you scrub the footage. Camera analytics turns the same feed into a forward-looking one. In a dining room, computer vision can tell whether a table has people at it, whether plates are gone and nobody has cleaned, how long a party has waited since sitting down, and how full the room is hour by hour. None of that needs new wiring: if a camera can produce an RTSP stream, software can read it.

What it can read on a restaurant floor

• Table state: available, occupied, needs attention, needs cleaning, reserved, and on TableWave Pro, a long sit flag for the manager.
• Timers: minutes since seated, since last staff visit, since the party left.
• Occupancy: how many tables are in use across the night, which is your real demand curve.

What it should never do in a restaurant: recognise faces, identify repeat guests, or score individual employees. Those features create legal exposure and poison staff trust, and no table-state problem requires them.

The privacy checklist to demand from any vendor

• Video processed on a computer inside your building, never uploaded to a vendor cloud.
• No facial recognition and no face templates, in writing.
• Faces blurred on any screen staff or owners see.
• Recording off by default, with a clear retention limit if you ever turn it on.
• A door notice for guests and a written notice for staff, supplied by the vendor in your language.

TableWave was built to pass this list: how we handle each point is on our privacy and security page.

What your cameras need to qualify

Less than you think. TableWave reads 640×360 video at two frames per second, which nearly any IP camera or recorder from the last decade can provide: Hikvision, Dahua, Reolink, UniFi, Axis, Tapo, EZVIZ or anything ONVIF. Cloud-only cameras that refuse to stream locally are the exception, and a replacement camera costs about $30 per zone. One computer in the building does the processing, so the system keeps running when the internet drops. Accuracy is measured in your own restaurant during the proof week, against what really happened on your floor, because lighting and angles differ too much between rooms for a universal claim. See pricing or book a 30-day pilot to have your camera layout checked in person.