The positive class is tiny
Weeks of footage may hold minutes of what matters.
Events, crowds and threat detection
AI Data for Security and Surveillance
Surveillance models are judged on rare events under poor conditions. We build datasets that reflect both.
Rare Events, Handled Properly
The events a security model exists to catch are, by definition, almost absent from the footage. Building a usable dataset means deliberate sampling, careful class definition and honest handling of the personal data involved — all three, or the programme fails on one of them.
Where Security Surveillance AI Programmes Struggle
Low light and low resolution
The hardest conditions are the normal ones.
Privacy is a hard constraint
Footage contains people who did not consent to be training data.
Behaviour is ambiguous
The line between unusual and suspicious has to be written down, not felt.
AI Use Cases We Support
Event and anomaly detection
Intrusion, loitering and abandoned-object labelling.
Crowd analytics
Density estimation and flow direction.
Perimeter monitoring
Line-crossing and zone-breach annotation.
PPE and safety compliance
Equipment detection in industrial settings.
Vehicle and access control
Entry, exit and dwell labelling at controlled points.
Forensic search sets
Attribute annotation for post-incident retrieval.
Data Types We Handle
- CCTV video Fixed-camera footage across light and weather conditions.
- Thermal and IR Night capture annotated alongside visible light.
- Access logs Structured events aligned to footage.
- Incident reports Operator records used as event ground truth.
How We Run the Programme
- Privacy first Faces and identifiers are minimised or masked wherever the task allows it.
- Event-window sampling Positives are found and expanded rather than waited for.
- Written behaviour definitions Ambiguous classes get an explicit rule and examples.
- Condition-stratified delivery Night, rain and low-resolution coverage is measured, not hoped for.
Build Security Surveillance AI on Data You Can Trust
Tell us what you are building and we will come back with a plan, a sample and a price.