How maiabboud-prog/multi-camera-people-tracking works
Multi Camera People Tracking is a configured, multi-camera people-detection and re-identification application. The user supplies a YAML configuration and camera sources; the runtime captures frames, detects people, extracts embeddings, and matches them against a bounded identity gallery shared across cameras. It annotates and combines camera views for optional display or video saving. Model downloads are a separate setup capability. The main loop is excerpted, but some camera setup and cleanup details are omitted, so the graph shows only evidenced runtime links.
Main entry points: main.py Tracking loop · download_model.py Model downloader · object_detection.py Person detector · feature_extraction.py Feature extractor
Components 13
Tracking runtime
- YAML configuration
main.py - Tracking loop
main.py - Camera sources
main.py - Identity gallery
main.py
Vision inference
- Person detector
object_detection.py - Feature extractor
feature_extraction.py
Output and display
- Box coordinate mapping
helpers.py - Camera view stacking
helpers.py - Live display
main.py - Video saving
main.py
Model setup
- Model downloader
download_model.py
Other components
- Operator
- Google Drive
Connections 13
- Operator → YAML configuration: provides settings
README.md - YAML configuration → Tracking loop: configures runtime
main.py - Operator → Camera sources: provides sources
main.py - Tracking loop → Camera sources: reads frames
main.py - Tracking loop → Person detector: predicts people
main.py - Tracking loop → Box coordinate mapping: rescales boxes
main.py - Tracking loop → Feature extractor: extracts features
main.py - Tracking loop → Identity gallery: matches and updates
main.py - Tracking loop → Camera view stacking: stacks views
main.py - Tracking loop → Live display: shows frames
main.py - Tracking loop → Video saving: writes frames
main.py - Operator → Model downloader: requests model setup
README.md - Model downloader → Google Drive: downloads model
download_model.py