ClearSight Edge

Your cameras already see the incident. ClearSight Edge tells someone while it is happening.

ClearSight Edge is video analytics that runs on a small NVIDIA Jetson computer next to your cameras. It watches for hygiene lapses, safety risks and signs of theft, and sends each incident to a person to review.

Talk to Full Stack See how it works

Video stays on site. Only incident records and short clips leave the device.

Illustration of a loading bay camera view with detections A warehouse loading bay seen from a ceiling camera. An amber dashed line marks a forklift zone on the floor. A forklift and two people have detection boxes. One person inside the forklift zone has a red box labelled "No hi-vis vest". The other person, outside the zone, has a cyan box and wears a vest. An incident notice at the bottom reads "PPE missing, severity high". Forklift zone Forklift, track 12 No hi-vis vest, track 7 Person, track 4 Camera 3, loading bay 14:02:37 SAST PPE missing, severity high In forklift zone for 1.5 s. Clip saved for review.
An illustration of the rule "person without a hi-vis vest in a forklift zone". The rule fires after the condition holds for 1.5 seconds.
Streams per device
4 to 8
Frame rate per stream
30 FPS
Capture to incident
Under 200 ms

These are design targets, not measured results. We measure them with you on your own cameras and hardware.

Most camera footage is watched after something has gone wrong.

A site with dozens of cameras rarely has someone watching every screen. Footage is searched after an injury, a failed audit or a stock count that does not add up. By then, the chance to step in has passed.

ClearSight Edge watches every stream for the specific things you care about. When it sees one, it records a short clip and alerts an operator within moments. The operator looks at the clip and decides what to do.

Every alert is a lead for a person to check, not a verdict. ClearSight Edge can miss events and can raise false alarms. That is why each incident has a confirm and dismiss step, with a record of who decided.

Three detection suites

Turn on the suites your site needs. Each rule works in zones that you draw on the camera picture, with timings that you set.

Hygiene

For kitchens, food production and clinical areas.

  • A person enters a prep area without using the hand-sanitiser station first.
  • Gloves, masks or hairnets are missing where your rules require them.
  • A surface was not cleaned within the time you set.

Health and safety

For warehouses, factories, yards and construction sites.

  • Hard hats, hi-vis vests or eye protection are missing in a zone that needs them.
  • A person enters a restricted area.
  • An emergency exit is blocked.
  • A person slips or falls.
  • A person lifts with an unsafe posture.

Loss prevention

For retail floors, stockrooms, loading bays and cash rooms.

  • Goods leave through the exit without a till (POS) scan on the linked lane.
  • Items are concealed.
  • A shelf is swept clear in bulk.
  • The loading bay has movement outside working hours.
  • A person enters the vault without a valid badge grant.

How it works at the edge

The analysis happens on site, on a device next to your cameras. Your cloud receives only small event records and short incident clips.

  1. Camera

    Your existing IP cameras send video over RTSP. USB webcams also work.

  2. Jetson device

    The device decodes each stream, finds objects such as people, PPE and products, follows them from frame to frame, and checks your rules.

  3. Operator console

    Operators see live video with detection boxes, review incidents and record a decision on each one.

  4. Your Azure tenant

    Event records go to your Azure IoT Hub. Incident clips go to your Azure Blob Storage.

What operators can do in the web console

Watch live video
Each camera shows in the browser with detection boxes drawn over the picture, in step with the video.
Draw zones
Draw a polygon on the camera picture to mark a prep area, forklift lane, exit or vault door. The device saves it.
Review incidents
Play the clip for each incident, then confirm or dismiss it. The console records who decided and when, as an audit trail.
Check device health
See frame rate and capture-to-result time for each camera. Cameras slower than 200 ms show in red.

Privacy and data handling

ClearSight Edge detects objects, PPE and behaviours inside the zones you draw. Video is analysed on the device at your site, and you decide what leaves it.

Kept on your site

  • All live video from every camera.
  • The analysis of every frame.
  • Zones, rules and timings.

Sent to your Azure tenant

  • Small event records: what rule fired, on which camera, when, and how severe.
  • Short video clips and snapshots of incidents.

Data goes to your own Azure subscription, not to Full Stack.

  • Network outages. The device stores event records and clips on its own disk during an outage, and sends them when the connection returns.
  • Predictable rules. The rules are deterministic. The same detections and settings always give the same incidents, so you can replay a case and see why an alert was raised.
  • Protected changes. Saving zones and recording decisions need an access token. Camera passwords are removed before addresses are shown in the console.

Deployment and hardware

One device covers one group of cameras. Add devices to cover more of the site.

DeviceNVIDIA Jetson, running JetPack 6.1
Cameras per device4 to 8 streams (design target)
Camera typesIP cameras over RTSP, and USB webcams
PackagingDocker containers on the device
ConsoleA web browser on your local network
CloudYour Azure IoT Hub and Azure Blob Storage
Other systemsTill (POS) scans and door badge grants, for the unpaid-exit and vault rules

Questions

Are the frame rate and speed figures proven?

No. 30 FPS per stream, 4 to 8 streams per device and under 200 ms from capture to incident are design targets. We measure them with you on your own cameras and hardware before you rely on them.

Will it catch every incident?

No system catches everything. ClearSight Edge can miss events and can raise false alarms. Every alert goes to a person who looks at the clip and confirms or dismisses it.

Do I need new cameras?

Usually not. ClearSight Edge works with IP cameras that stream over RTSP, and with USB webcams. Camera position matters: the picture must show the zone you want to watch.

What happens when the internet connection drops?

Detection and alerts in the console continue, because they run on site. The device queues event records and clips, and sends them to Azure when the connection returns.

Where does our data go?

Live video does not leave your site. Event records and short incident clips go to Azure IoT Hub and Azure Blob Storage in your own Azure tenant.

Why did it raise an alert?

Each rule is deterministic. A rule fires only when its condition holds for a set time, for example a person without a vest in a forklift zone for 1.5 seconds. You can replay the same detections and get the same result.

Find out if ClearSight Edge fits your site

Tell us what you want to detect and which cameras you have. We will tell you what is possible.

Contact Full Stack