Argus

multi-camera surveillance ingestion & correlation engine

argus is built on top of the RT-correlation vision pipeline. it takes direct feeds from multiple localized IP surveillance cameras, ingests them asynchronously via background workers, and applies real-time object tracking models locally.

instead of just storing footage like traditional CCTV systems, it evaluates movements, calculates path histories across camera positions, and creates unified logs showing trajectory timelines.

// telemetry data

feeder count: 3 streams
frame ingestion: 30 FPS
latency: ~80ms local
storage: local sqlite logs

// components

ingest: FFmpeg + RTSP
inference: YOLO-TensorRT
tracking: SORT Kalman
dashboard: NextJS / React

[ FEED VISUALIZER NOT RENDERED - EXTERNAL DISPLAY LOCK ]

video interfaces are configured for local system monitors. web panels show structural stats only.