The Securade.ai Edge App is what runs on the box next to your CCTV cameras and turns the video into safety enforcement. This is a walk-through of what's actually in the app, how the configuration works, and what comes out the other end. If you're evaluating it or about to deploy, this is the tour.

What the Edge App is

The Edge App is the on-site component that ingests camera feeds, runs the AI models, applies your safety policies, and pushes alerts when violations happen. It's designed to be configured by a safety officer or supervisor, not by an ML engineer. The admin interface handles the camera setup, the policy definitions, and the alerting channels.

What's in the app

  1. Admin interface. The entry point. Where you configure feeds, write policies, and manage the deployment.
  2. Camera feed integration. Live preview of connected RTSP feeds so you can confirm what each camera is seeing before you write policies against it.
  3. Policy configuration. The rules that determine what counts as a violation. Three families to start with:
    • PPE detection (hard hats, vests, gloves, etc.)
    • Proximity detection (people too close to moving equipment)
    • Zone management (restricted areas)

Zone management

Zone management lets you draw polygons on the camera view to mark areas where people shouldn't be. When the model sees a person inside a defined zone, that's a violation and an alert fires. This is what catches the "someone walked into the forklift lane" class of incident.

Applying policies to a live feed

Once a policy is defined, you load it against a feed and the app overlays the policy on the live camera view. You can watch the detections in real time. This is how you confirm the policy is doing what you wanted before you walk away and trust it to run unattended.

Settings worth knowing about

  1. Model selection. Different models for different workloads. Pick the one that matches the site.
  2. Confidence threshold. Tune how sensitive the detector is. Higher threshold means fewer alerts but more missed events; lower means the opposite.
  3. Privacy. Face masking is built in for sites where you can't store identifiable footage.
  4. Alert channels. Telegram is the default; other channels including webhook and email are configurable.

When the configuration looks right, start the server. From that point the app is running continuously, scoring frames, and pushing alerts when policies are violated.

Reporting

The dashboard is where you go to see what's been happening. It aggregates violations by camera, by policy, by date, by shift. The point isn't to micromanage individual events; it's to see patterns over weeks and months and use those patterns to drive training and process changes.

  • Date-wise violation tracking. Which days, which shifts, which areas of the site.
  • Trend analysis. Whether the numbers are getting better or worse, and where to focus the next safety intervention.

That's the tour. The Edge App is the working end of the platform, the thing that actually watches the camera and decides whether to send an alert. It's designed to be configurable by a safety team, not a data science team, and to run for months without intervention once it's set up. If you want to evaluate it on your own footage, the source is open and the docs walk you through a first deployment.