How quoccuong66/project-monitoring-model works
This is an edge-based camera monitoring system for detecting falls and other observable abnormal behavior, escalating risk for human verification rather than diagnosis. Its primary workflow ingests live or recorded video, tracks people, extracts pose and temporal behavior signals, fuses risk, and dispatches alerts; a FastAPI service exposes monitoring controls. The README also documents model training. Runtime links follow the README and shown imports/references; unsampled implementation details and some wiring are omitted rather than inferred.
Main entry points: realtime_pipeline.py Inference pipeline · server.py Monitoring API · risk_engine.py Risk fusion · alert_manager.py Alert management
Components 24
Video ingestion
- Frame stream
src/camera/stream.py - Camera sources
src/camera
Perception and reasoning
- Inference pipeline
src/pipeline/realtime_pipeline.py - Motion gating
src/detection/motion_gater.py - Person detection
src/detection/detector.py - Person tracking
src/tracking/track_manager.py - Pose extraction
src/pose/keypoints.py - Temporal features
src/temporal/temporal_features.py - Action classification
src/action/action_classifier.py - Anomaly scoring
src/anomaly/anomaly_score.py
Risk and alerts
- Risk fusion
src/risk/risk_engine.py - Risk thresholds
src/risk/threshold_manager.py - Alert management
src/alerts/alert_manager.py - Event logging
src/alerts/event_logger.py - Notification dispatch
src/alerts/notification.py - Alert channels
Operator interfaces
- Application launcher
main.py - Monitoring API
src/api/server.py - Web dashboard
frontend/app.js
Model training
- Dataset preparation
training/prepare_ntu.py - Action model training
training/train_action.py - Anomaly model training
training/train_anomaly.py
Other components
- Care operator
- Camera or video
Connections 24
- Camera or video → Camera sources: provides video
README.md - Camera sources → Frame stream: produces frames
src/camera/webcam.py - Application launcher → Inference pipeline: starts inference
main.py - Application launcher → Monitoring API: starts service
main.py - Monitoring API → Camera sources: selects source
src/api/server.py - Monitoring API → Inference pipeline: starts pipeline
src/api/server.py - Web dashboard → Monitoring API: uses monitoring service
README.md - Frame stream → Inference pipeline: supplies frames
README.md - Inference pipeline → Motion gating: gates inference
src/pipeline/realtime_pipeline.py - Inference pipeline → Person detection: detects people
src/pipeline/realtime_pipeline.py - Inference pipeline → Person tracking: updates tracks
src/pipeline/realtime_pipeline.py - Inference pipeline → Pose extraction: extracts pose
src/pipeline/realtime_pipeline.py - Inference pipeline → Temporal features: builds features
src/pipeline/realtime_pipeline.py - Inference pipeline → Action classification: classifies actions
src/pipeline/realtime_pipeline.py - Inference pipeline → Anomaly scoring: scores anomalies
src/pipeline/realtime_pipeline.py - Inference pipeline → Risk fusion: assesses risk
src/pipeline/realtime_pipeline.py - Risk fusion → Risk thresholds: loads thresholds
src/risk/risk_engine.py - Risk fusion → Alert management: provides assessment
README.md - Alert management → Event logging: logs event
src/alerts/alert_manager.py - Alert management → Notification dispatch: dispatches event
src/alerts/alert_manager.py - Notification dispatch → Alert channels: sends alerts
README.md - Care operator → Web dashboard: reviews status
README.md - Dataset preparation → Action model training: prepares data
README.md - Dataset preparation → Anomaly model training: prepares data
README.md