Nexma InsightEngine
Find every object, track every move
Nexma InsightEngine is the platform's perception layer — production-grade computer vision for overhead imagery, full-motion video, and live sensors. It detects and classifies stationary and moving objects with high accuracy, tracks each one across frames, and writes every result straight into the world model. Imagery, video, and sensor streams arrive faster than any team can watch them; InsightEngine turns that flood into typed, trackable entities.
Core concepts
A bounding box in a viewer is not actionable. InsightEngine exists to make perception part of the world model, not a separate silo.
- Detect and classify objects — vehicles, vessels, aircraft, infrastructure — across overhead and full-motion imagery at scale.
- Track over time. Maintain object identity across frames and passes, so objects are followed through time rather than re-detected as strangers.
- Fuse many sensors into one operational picture — EO, SAR, video, and ground sensors, deduplicated and georeferenced.
- Write to the world model. Results land in the DataStore as typed entities, so a detection can trigger a workflow rather than sit in a viewer.
A detection that lives in the world model can launch an agentic workflow — alert, task, or dispatch — automatically. A detection in a standalone viewer cannot.
How it works
Point InsightEngine at any overhead or motion feed. It perceives, tracks, fuses, and grounds — turning raw pixels into entities the rest of the platform understands.
1Feed: full-motion video over an area of interest
2 1. Detect classify every object in frame, with confidence
3 2. Track lock a track onto each one; hold identity across frames
4 3. Fuse merge with EO, SAR, and ground-sensor hits → one picture
5 4. Ground write typed entities to the DataStore (model + confidence + source)
6 5. Act a detection can trigger an AgentEngine workflowPerception across modalities
The engine works across overhead and motion imagery and ground sensors at once, fusing everything into a single georeferenced picture.
| Stage | What it covers |
|---|---|
| Perceive | Object detection across EO, SAR, and full-motion video; change detection; multi-modal input |
| Track and fuse | Temporal tracking across frames and passes; deduplicated, georeferenced sensor fusion; confidence and provenance on every hit |
| Ground and act | Typed entities written to the DataStore; workflow triggers; full history retained |
Change detection
Compare imagery across time and InsightEngine surfaces what is different — new construction, movement, encroachment — automatically, with no frame-by-frame review. Combined with tracking, this gives the operational picture a memory: not just what is in the current frame, but what has appeared, moved, or changed since the last pass.
Fused, grounded, and persistent
Detections from every sensor land on the same map, deduplicated and georeferenced, so teams reason from one shared truth instead of a stack of disconnected viewers. Each detection carries its model, confidence, and source for trust and review. Because results persist with time — tracked through the DataBase — the picture has continuity, and a detection can immediately feed analysis, alerting, or dispatch.
Models watch imagery and video continuously, so events surface automatically instead of waiting in an analyst's backlog.
Where to go next
- Nexma SatelliteEngine — task fresh imagery for InsightEngine to analyze.
- Detection capability — what perception enables end to end.
- Nexma AgentEngine — turn a detection into an automated response.
- DataBase — track detected objects through time at scale.