Skip to main content

FOVEA

Flexible Ontology Visual Event Analyzer

Build video annotation ontologies that capture how different analysts interpret the same events. Multiple perspectives, shared reality, semantic grounding.

Get StartedDocumentation

What Makes FOVEA Different

🔗

Types vs. Instances

Separate analyst interpretations (EntityType, EventType) from shared world objects (Entity, Event). Different analysts assign different types to the same real-world objects.

👥

Persona-Based Ontologies

Each analyst maintains their own ontology with distinct entity types, event types, and role definitions. Capture tactical, operational, and strategic perspectives on the same video.

🌐

Semantic Grounding

Link types to Wikidata entities for semantic interoperability. Import definitions, GPS coordinates, and temporal data directly from external knowledge bases.

🔀

Collections & Relations

Group entities and events into semantic collections. Define custom relation types to capture organizational hierarchies, causal links, and temporal patterns.

⏱️

Rich Temporal Model

Express vague time (early morning, late 2024), deictic references (yesterday, next week), cyclical patterns, and temporal relationships across multiple videos.

🎬

Advanced Keyframes

Keyframe-based annotation with linear, bezier, ease-in-out interpolation. Toggle visibility for occlusion handling. AI-powered tracking acceleration.

When Different Perspectives Matter

FOVEA is designed for situations where multiple analysts need to interpret the same video footage through different analytical lenses, while maintaining a shared understanding of underlying reality.

  • Intelligence analysis with tactical, operational, and strategic viewpoints
  • Disaster response coordination across emergency services and NGOs
  • Clinical research where radiologists and specialists interpret the same scans
  • Legal proceedings with prosecution, defense, and expert witness perspectives
  • Ecological monitoring where biologists track different interaction types
  • Ethnographic research capturing participant and observer interpretations

Core Concepts

  1. Create personas representing different analytical roles
  2. Build persona-specific ontologies with entity and event types
  3. Annotate video linking regions to types OR shared world objects
  4. Ground types semantically using Wikidata references
  5. Define collections and relations to capture structure
  6. Export with conflict resolution for collaborative workflows
Example: A disaster analyst sees "Infrastructure Damage" while a humanitarian sees "Shelter Need" on the same collapsed building instance.

Build ontologies that capture how analysts think

Open source, Docker-based deployment, runs on CPU or GPU

Get StartedView on GitHub