INVICTUS

PROBLEM: - Aerial drones capture hours of high-resolution video during floods, earthquakes, and collapses. - Exhausted rescue teams must manually scan empty terrain frame-by-frame to locate survivors. - Valuable time is lost during the critical golden hour of disaster response. SOLUTION: - Project Triage is an AI-assisted aerial reconnaissance dashboard built for disaster response teams. - It ingests raw drone video and field imagery, discarding empty footage and isolating prioritized survivor candidates. HOW IT WORKS: - Keyframe Extraction: Automatically filters out thousands of redundant, empty frames using visual change dynamics. - Spatial Segmentation via Meta SAM 3.1: Isolates detected targets and survivor contours with pixel-accurate spatial masks. - Contextual Analysis via Meta Muse Spark 1.3: Evaluates environmental hazards (such as rising floodwaters and structural debris) to generate prioritized triage ratings (P1 Critical to P4 Monitor). - Zero-Dismissal Safety Protocol: Ensures ambiguous signals are always escalated to human operators rather than silently dropped. - Tactical Mission Dashboard: Delivers candidate telemetry, visual crops, and hazard assessments in a single pane for instant dispatch decisions.