We mark up your images and video, saying what each thing is and where, so AI can learn to recognise it.
Explore serviceData & annotation
Know what happens in your video, as it happens.
Video holds more information than a photo, and it is harder to work with. We label what happens across the frames, the objects, the actions and the events, and we build systems that watch recorded footage or live cameras for you. Every project starts by agreeing exactly what counts as an event, so the system flags what your team actually cares about.
A real example
What a model sees once it has learned from labelled video.
Veesion's software watches a shop's existing cameras and alerts staff to the gestures that tend to come before theft. To do that, it first had to learn from video that people labelled frame by frame. Innovaith labels video like this for Veesion.
Clip and image © Veesion, shown with their permission.
Try it yourself
The kinds of label, one at a time.
Every frame of a video is labelled like this. Switch between the layers to see the kinds of label we add.
Boxes. A box around every person and object that matters, each with a name. This teaches a model what is in the picture and where.
Our own illustration, drawn on a stock photo so you can switch each kind of label on and off. It is not client footage.
What we offer
Video Detection services.
- 01
Object detection and tracking
Objects detected in each frame and followed with persistent IDs as they move, overlap and leave the scene.
- 02
Action recognition
Clips and segments labelled with the activity taking place, from simple movements to multi-step actions.
- 03
Event and anomaly detection
Rules and models that flag unusual behaviour, safety incidents or process deviations in footage.
- 04
Temporal annotation
Frame-level labels with start and end times, so models learn when something happens as well as what.
- 05
Multi-camera tracking
Consistent identities for people and objects as they move between cameras with different angles and fields of view.
- 06
Real-time alerting
Detection pipelines that process live streams and notify your team or systems when a defined event occurs.
How we work
From guidelines to delivered data.
AI takes the first pass, trained people review it, and quality is checked before delivery.
- 01
Scope
We agree label definitions, edge cases, formats and quality targets with you.
- 02
Pilot
A small batch tests the guidelines before any volume work starts.
- 03
Annotate
Trained annotators label to the agreed guidelines, with tooling to match.
- 04
Review
Multi-stage quality review against the specification, with feedback loops.
- 05
Deliver
Data in your format, with notes on what was checked, then we iterate.
Where it helps
Typical projects.
The kinds of work this service usually supports.
Workplace safety
Detecting missing protective equipment, restricted-zone entry or unsafe movement on sites and factory floors.
Retail and footfall
Understanding queues, dwell time and movement through a store using existing camera footage.
Traffic and transport
Counting and classifying vehicles, and spotting incidents or congestion on roads and in depots.
Production line monitoring
Watching production lines for defects, jams or missed steps as items move past a camera.
Details
Output formats
- COCO JSON (video)
- MOT Challenge format
- CVAT video XML
- YOLO TXT
- Temporal segments (JSON)
- Frame-level CSV
- Alert payloads (JSON via webhook)
- Custom schemas
Questions
Can this run on our existing cameras?
Often, yes. We review camera placement, resolution and frame rate early and tell you plainly if any of them will limit what the system can detect.
Does our footage have to leave our premises?
Not necessarily. Detection can run on edge devices or in your own cloud account, and annotation can be carried out under NDA with access controls agreed with you.
How do you handle privacy in video data?
We agree data handling rules before work starts, including who can view footage, how long it is kept and whether faces or number plates should be blurred.
More data & annotation services
All services
Match footage turned into data: who was where, what happened and when, for coaches, scouts and media teams.
Explore service
People who review, rank and correct what an AI writes, so it learns the tone and judgement you want.
Explore service
Idea → Architecture → Prototype → Production → Scale
Talk to us about Video Detection.
Tell us what you're trying to build, test, automate or improve. We'll tell you honestly where we can make a real difference, and where we can't.