Multi-sensor data fusion
Imagery, signals and reports combined into one picture with provenance for analysts.
The problem
Analysts spend most of their time collecting and reconciling sources instead of assessing them.
Data we need
- Imagery and video feeds
- Signal and track data
- Text reports
How it works
- Agree sources, formats and classification rules
- Build ingestion on the accredited network
- Train entity matching and change detection models
- Present fused results with source links
- Log every model output for review
What it changes
Targets are agreed with you before we build, and measured on your data before release.
- More time on assessment, less on collation
- Every conclusion traceable to its sources
- Consistent picture across teams
Deep learning
Neural networks for problems where hand-written rules and classic models fall short.
- CNNs
- Transformers
- Autoencoders