Video and camera ingestion
Inspection sessions take frames from a USB camera on an Android device, a phone camera or a browser webcam — roughly one frame a second from Android. Recorded video and images can be used for clearly labelled simulations.
Platform overview
One workflow from camera frame to closed-out action
VISION is a module of Green Codes Pro with its own AI inference service. It takes frames from inspection cameras, raises potential hazards, routes them to people for review, and keeps the evidence behind every decision.
AI-generated footage (Adobe Firefly) · illustrative
Architecture
How the pieces fit together
Detection runs on a server, not on the camera or phone. Only the VISION module talks to the inference service.
Camera sources
- USB camera on Android
- Phone camera
- Browser webcam
- Recorded video (simulation)
VISION module in Green Codes Pro
- Sessions & rules
- Evidence store
- Review queue
- Dashboards & reports
VISION inference service
- Object detection
- Tracking across frames
- Duplicate suppression
Reviewers
- Site HSE
- Head Office
Green Codes Pro records
- Observations
- Corrective actions
Components
What each part does
AI analytics
A self-hosted inference service runs the detection model, follows objects across frames, waits for a detection to persist before raising it, and suppresses duplicates.
Rules
Each project sets which hazards apply, their severity, and — on a fixed camera — the exclusion zones a supervisor draws on the image.
Review and verification
Every finding starts as a potential hazard. A competent person confirms it or dismisses it with a reason such as “false positive” or “already controlled”.
Event evidence
Original and annotated frames are kept only when there is an event — VISION does not record continuously. Head Office reviews can attach short video clips.
Oversight dashboards
Control room, sessions, review queue and analytics, plus a Head Office view with safety walk reports, a findings register, a 5×5 risk matrix and escalation.
Alerts
The right people hear about it — in the right way
-
In-app and e-mail notifications
High-severity hazards, escalations, assignments and completed safety walk reports.
-
Spoken alerts on the inspector’s device
Arabic or English voice warnings through a Bluetooth earpiece, so the inspector never has to look at a screen.
-
Head Office escalation
Findings can be escalated to the people who need to act on them, with the evidence attached.
Integration boundaries
What VISION does — and what it deliberately does not
VISION does
- Raise potential hazards from camera frames
- Keep the evidence behind each finding
- Route findings to people for confirmation
- Create Green Codes Pro observation drafts from confirmed findings
VISION does not
- Decide on its own that a violation happened
- Identify workers or use facial recognition
- Record video continuously
- Stop machinery, issue permits or replace supervision
Deployment
How VISION is deployed
Inside Green Codes Pro
VISION is enabled per organisation within Green Codes Pro, using Pro’s users, roles and projects.
Self-hosted inference
The inference service is containerised and runs on a CPU or an NVIDIA GPU server. Each service runs one model; hosting arrangements are agreed per project.
Separate simulation
Simulation sessions and exhibition demonstrations never write to live records.
See VISION with your own site in mind
Book a guided demonstration. We will walk through the exhibition scenarios, the human review workflow, and what a rollout on your projects would involve.