Security & surveillance
Every alert checked by a person before anyone is sent out.
Annotation and live review for perimeter, site and port monitoring, with privacy redaction built in.
From the camera to the guard on shift
Every part of the workflow where we can take work off your team.
Detection
Tracking
Alerts
Privacy and testing
We work with CCTV, thermal, drone and night vision.
What we label
Every class is written into a guide with your team before the first batch. Files come back as COCO JSON, YOLO, Pascal VOC, CSV, JSONL or your own schema.
Camera footage
People and vehicles
Bounding box and track
Person, car, truck, bike, boat
Behaviour
Time span and class
Loitering, climbing, running, tailgating
Zones
Polygon
Restricted area, fence line, gate
Alerts
Alarm review
Class and reason
Real intrusion, animal, weather, shadow, staff
Privacy
Redaction
Bounding box and blur mask
Face, number plate, screen
How we test the model
The tests that matter for security & surveillance models, and what you get back from each.
False alarm rate
Every reviewed alarm recorded as real or false, with the reason.
You get backFalse alarms by camera, hour and cause
Night and weather recall
Detections checked on purpose at night, in rain, fog and glare.
You get backRecall split by condition
Tracking accuracy
IDs checked across cameras for switches and losses.
You get backID switches and lost tracks by scene
Where security & surveillance teams start
Problems teams bring us, and the first piece of work we would scope.
Guards are sent out for shadows, animals and rain.
What we doEach alarm is confirmed or dismissed by a person in under 2 minutes, around the clock.Live reviewThe model misses people at night and in bad weather.
What we doFootage from those conditions collected and labelled as its own set.AnnotateFootage cannot leave your network with faces in it.
What we doFaces and plates redacted before anything is shared or labelled.AnnotateRelated work
Defect annotation on building imagery in CVAT, with a webhook loop between annotators and reviewers.
50,000 facade annotations a week for building condition assessments
Part identification and damage segmentation for insurance claims AI.
200,000 polygons for automated vehicle damage assessment
2,000 robot videos described every week by a team built in seven days
Half a million images for autonomous navigation, with every file kept on EU servers
Client and volumes [REAL DATA: Work that can be described publicly]
Bring us the security & surveillance data your model gets wrong.
Send a small batch and get it back done, free, usually within 24 to 48 hours. You judge the work before you pay for anything.