SeraFina is designed around four model families: Visual, Audio, Predict, and Analyze. Each has a different role within a modular architecture intended to connect information with the needs of the operator.
Understanding incoming information
Visual focuses on visual perception, while Audio handles audio interpretation. Together, these model families are intended to help turn different types of incoming information into material that can support further analysis.
Moving from inputs to assessment
Predict addresses predictive assessment: considering observations and historical information to assess potential hazards and gaps in the available picture. Analyze focuses on objective-based analysis, interpreting information in relation to mission priorities and user requirements.
The four families describe complementary responsibilities rather than a single model expected to handle every task. A modular approach is intended to make it easier to adapt capabilities and integrations to different use cases.
A development framework
This architecture provides a framework for development. The performance of individual models and the combined system will need to be tested against representative data and operational requirements. Outputs are intended to support trained personnel as they assess situations and choose a course of action.