How leo4bits/fire_smoke_person_detection works
This is a real-time fire/smoke and air-quality monitoring application. Its main workflow captures webcam frames, enhances them, detects objects, and overlays detections alongside sensor risk predictions. In parallel, a serial-reader thread parses ESP32 readings and obtains predictions from a Random Forest model. The README was empty, so capabilities and relationships here are based on the sampled runtime source; model artifact wiring is not shown clearly enough to connect the repository files.
Main entry points: mainTest.py Live monitor · thread.py Serial reader · imgPreProcessing.py CLAHE enhancement · MODEL_USING.py YOLO inference
Components 11
Vision detection
- Live monitor
mainTest.py - CLAHE enhancement
PreProcessing/imgPreProcessing.py - YOLO inference
MODEL_USING.py - YOLO weights
best.pt
Sensor monitoring
- Serial reader
threadGetData/thread.py - Sensor prediction
PreProcessing/numPreProcessing.py - Random Forest weights
random_forest_air_quality.pkl
Live display
- Label formatting
PreProcessing/outputGUI.py - Annotated frame
mainTest.py
Other components
- Webcam
- ESP32 sensors
Connections 9
- Webcam → Live monitor: provides frames
mainTest.py - Live monitor → CLAHE enhancement: enhances frames
mainTest.py - Live monitor → YOLO inference: requests boxes
mainTest.py - Live monitor → Serial reader: starts thread
mainTest.py - ESP32 sensors → Serial reader: sends readings
threadGetData/thread.py - Serial reader → Sensor prediction: processes readings
threadGetData/thread.py - Serial reader → Live monitor: shares predictions
mainTest.py - Live monitor → Label formatting: formats labels
mainTest.py - Live monitor → Annotated frame: draws overlays
mainTest.py