EVUbeam by EVUNITY

Technology

Built to decide, and to show why

EVUbeam is a chain of small, separate steps. Each does one job, and each leaves a record of what it decided. That is how a team can trust an alert, and check it months later.

Five steps, one pipeline

Every picture from your cameras passes through the same five steps, in order.

  1. Capture

    Reads live streams from your recorder and keeps a short rolling buffer of frames, so every alert can show what happened just before it.

  2. Detect and track

    Finds people and objects in the picture and follows each one from frame to frame, so one person stays one person.

  3. Measure

    Measures how people move and how things are handled: contact, force, movement, repetition and body pose.

  4. Judge

    Applies your rules and a small motion-recognition model. Independent signals must agree before anything becomes an alert.

  5. Record

    Writes every alert, with its picture, clip and written account, to a record that is added to and never rewritten.

Three separate decision paths

The same camera answers three different questions. Each is decided on its own terms and they never vote on each other. If two paths name the same person in the same second, they merge into one incident that keeps the higher priority.

Intrusion

Is somebody inside an area you marked, and for how long? Decided by the shape of the area, the waiting time and a short confirmation window.

Handling of goods

Was something touched, struck, moved or damaged? Decided by contact, strength of movement, displacement and repetition.

Person action

What is the person doing, and is it safe to raise an alert on it? A motion model names the action; body-pose measurement and a final check decide whether it becomes an alert.

Runs on hardware you choose

EVUbeam is installed as a Windows application on a machine you own. Analysis runs on a graphics card in that machine, so no picture has to leave the building.

Your own machine

Windows 10 or 11 (64-bit). An NVIDIA graphics card for live analysis; the console itself runs on an ordinary processor.

Your existing cameras

Connects to your recorder over standard network video protocols. Phones on your network can act as extra cameras.

No internet needed

The models run locally. If the internet goes down, EVUbeam keeps watching and keeps its records.

It finds the recorder again

An address is where a recorder can be reached. It is not what the recorder is. EVUbeam keeps the two apart, so a recorder whose network address changes is found again without anyone typing a new one.

  • The recorder is recognised by its own identity, found with the network's standard discovery, never by scanning for addresses.
  • If a different device takes over the old address, EVUbeam does not connect to it. The camera stays honestly offline until an administrator confirms.
  • Discovery stays on your local network and never uses a password.

Open to your other systems

The console is a web application on top of an API that handles sign-in, permissions, alerts, evidence export and recorder playback. Talk to us about connecting EVUbeam to the systems you already use.

Security built in

  • Passwords are stored as salted hashes. Sessions are signed and can be ended per account.
  • One permission table covers every action. A guard, a supervisor and a manager see and can do different things.
  • An append-only audit log records who changed what, who watched which footage, and who exported evidence.
  • The recorder's password stays on the machine. It is never sent to a browser, never shown on screen and never written to a log.
  • The desktop app is sandboxed and only talks to its own console.
  • Automated tests include hostile input and permission-boundary checks, and a scan makes sure no password is committed to the code.

What we measure, and what we do not claim

The motion-recognition model is small and decides in about a millisecond on an ordinary processor. We test EVUbeam as installed, on real cameras, and we do not quote an accuracy percentage until it has been measured on your site with your cameras and your lighting.

Faces must be visible for a person to be recognised. If they are not, the result is "not recognised", never "intruder".

Put your cameras to work.

Tell us about your site. We will show you what EVUbeam would watch for, on the cameras you already have.

Request a demo