Questions buyers ask us

Yes. VISION runs as a module of Green Codes Pro with AI detection, human review, evidence, Head Office reporting and the Pro observation workflow. Every engagement starts with a guided demonstration.

Yes. VISION runs as a module of Green Codes Pro and uses its users, projects, observations and corrective actions. If you do not use Pro yet, we can talk about both together.

Missing safety helmets, missing high-visibility vests, exclusion-zone entry on fixed views, people near excavators, dump trucks and bulldozers, and visible smoke. Fall hazards and other site conditions are available as simulation scenarios for training. Details are on the AI safety detection page.

Accuracy depends on the detector and on site conditions such as camera position, lighting and distance, so we do not publish a single headline figure. No AI detector is perfect, which is why every finding is reviewed. We walk you through detection performance on footage like yours during a demonstration.

Every finding is reviewed. A reviewer dismisses a wrong finding with a reason — false positive, duplicate, not applicable, already controlled, poor image or other — and the reason stays on the record, which shows where a detector needs improvement. Uncertain findings are never spoken as alerts or reported automatically.

No. VISION never makes an automatic violation decision. A competent person confirms every finding, and Head Office approval keeps separation of duties.

A USB (UVC) camera connected to an Android device, the Android device’s own camera, and browser webcams. Recorded video can be used for simulation and training. Ask us about a specific camera model.

Yes. Tell us about your camera estate — models, recording system and network — and we will recommend the right VISION set-up for your sites.

The inspector kit uses a lightweight USB camera with the Android device carried in a vest or shoulder pouch, not on the helmet. On the helmet, the camera only uses an attachment approved by the helmet manufacturer — we never drill or glue helmets — and chest or vest mounts are also available.

Not with the standard kit. The camera hardware is not rated for explosive atmospheres (ATEX / IECEx) and must not be used in classified hazardous areas.

On the VISION inference service, a containerised server component running on a CPU or an NVIDIA GPU. Devices send frames to it; they do not run AI models themselves. Hosting arrangements are agreed per project.

Live detection needs a connection to the inference server. The Android app keeps captured evidence encrypted in an offline queue and uploads it when the connection returns.

It depends on your sites, hardware and how many reviewers are involved. We scope that with you after a discovery call and a demonstration rather than promise a fixed timeline.

No. There is no facial recognition, no worker identification and no per-worker statistics.

No. VISION keeps frames only when they carry an event, plus short clips attached for Head Office review. Retention is configurable from 7 days to 5 years, with 90 days as the default.

VISION is built around data minimisation: no identification, event-only evidence, private storage and configurable retention. We work through your Saudi Personal Data Protection Law (PDPL) requirements with your data protection team.

The exhibition scenarios — missing helmet, missing high-visibility vest and several PPE gaps — run through VISION’s detection, the review workflow, Head Office reporting and the Green Codes Pro observation flow.

No. Demonstration footage is AI-generated with Adobe Firefly so no real worker is filmed. It is labelled as such on screen. The detections you see on it are the model’s real output.

Yes. The web module, the Android app, reports, summaries and spoken alerts all work in Arabic and English.