How sanjaykraj37-cloud/fire-camera-detector works
Fire Camera Detector is a local fire-warning application with two monitoring paths: a colour-and-motion heuristic and a trained whole-frame EfficientNet classifier. Operators provide webcam or video input; frames are analysed, candidates are confirmed across consecutive frames, and alerts produce visual output, sounds, snapshots, and (for AI monitoring) CSV records. A separate workflow downloads folder-labelled data, trains and evaluates the classifier, and saves reusable model artifacts. The graph reflects sampled Python source and README capabilities; generated detection files are runtime artifacts.
Components 22
Live Detection
- Heuristic Monitor
fire_detector.py - Flame Masking
fire_detector.py - Region Filtering
fire_detector.py - Heuristic Confirmation
fire_detector.py - Heuristic Overlay
fire_detector.py
AI Monitoring
- AI Monitor
ai_fire_detector.py - Model Loading
ai_fire_detector.py - Fire Classifier
ai_fire_detector.py - Display Localization
ai_fire_detector.py - AI Confirmation
ai_fire_detector.py
Model Training
- Dataset Downloader
download_dataset.py - Training Runner
train_model.py - Stratified Splits
train_model.py - Classifier Training
train_model.py
Alert Artifacts
- Model Checkpoint
models/fire_classifier.pt - Evaluation Metrics
models/metrics.json - Alert Snapshots
- Alert CSV Log
- Alarm Output
alerm_audio.wav
Other components
- Operator
- Camera Or Video
- Kaggle Dataset
Connections 26
- Operator → Heuristic Monitor: starts monitor
- Camera Or Video → Heuristic Monitor: supplies frames
- Heuristic Monitor → Flame Masking: calls masking
- Heuristic Monitor → Region Filtering: filters regions
- Heuristic Monitor → Heuristic Confirmation: confirms frames
- Heuristic Monitor → Heuristic Overlay: draws overlay
- Heuristic Monitor → Alert Snapshots: writes snapshots
- Heuristic Monitor → Alarm Output: plays alert
- Operator → AI Monitor: starts monitor
- Camera Or Video → AI Monitor: supplies frames
- AI Monitor → Model Loading: loads model
- Model Loading → Model Checkpoint: reads checkpoint
- AI Monitor → Fire Classifier: predicts confidence
- AI Monitor → Display Localization: localizes display
- AI Monitor → AI Confirmation: confirms frames
- AI Monitor → Alert Snapshots: writes snapshots
- AI Monitor → Alert CSV Log: appends events
- AI Monitor → Alarm Output: plays alert
- Operator → Dataset Downloader: requests download
- Dataset Downloader → Kaggle Dataset: downloads data
- Operator → Training Runner: starts training
- Training Runner → Stratified Splits: creates splits
- Training Runner → Classifier Training: trains classifier
- Training Runner → Model Checkpoint: saves best model
- Training Runner → Evaluation Metrics: writes metrics
- Dataset Downloader → Training Runner: provides path