AI/ML enhancements
Longer-term intelligence initiatives layered on top of the core visualization and alerting experiences.
Phase 4 – Automated highlights
- Train CV models to flag rapid cloud-top cooling, eye-like structures, and overshooting tops.
- Surface detected events as timeline markers and contextual callouts.
Phase 5 – Motion & anomaly analytics
- Run optical-flow over successive frames to derive wind vectors and projected storm paths.
- Watch for pixel-domain anomalies (fog, volcanic ash, wildfire smoke) and trigger alerts.
Phase 6 – Conversational intelligence
- Fine-tune an LLM on EMWIN bulletins to normalize jargon and answer user questions.
- Fuse imagery + text embeddings for cross-modal search such as “latest SST anomaly near Baja.”