How ylaxor/flowar works
FlowAR is a Streamlit research framework for developing and evaluating binary-sensor human activity recognition. Users load a smart-home dataset, explore it, configure segmentation and representation, and run classification or hyperparameter search; cross-validation evaluates the resulting pipeline. The graph preserves the documented GUI stages and backend evaluation capabilities. Some page implementations and algorithm variants were not sampled, so their internal calls are omitted rather than inferred.
Main entry points: cv.py Classification cross-validation · main.py Streamlit app · 3_classification.py Classification page · data_loader.py Dataset loader
Components 20
Research interface
- Streamlit app
gui/main.py - Dataset exploration
gui/pages/0_exploration.py - Segmentation page
gui/pages/1_segmentation.py - Representation page
gui/pages/2_representation.py - Classification page
gui/pages/3_classification.py - Method controls
gui/components/shared.py - Dataset loader
gui/components/data_loader.py
Dataset and domain
- Smart-home dataset
backend/dataset/methods/base.py - Events and segments
backend/models.py
Recognition pipeline
- Segmentation methods
backend/segmentation/methods - Feature extractors
backend/representation/methods - Activity annotators
backend/annotation/methods - Classifiers
backend/classification/methods
Evaluation and tuning
- Segmentation metrics
backend/segmentation/quality/metrics.py - Segmentation validation
backend/segmentation/quality/cv.py - Representation analysis
backend/representation/quality - Classification cross-validation
backend/classification/quality/cv.py - Classification metrics
backend/classification/quality/metrics.py - Hyperparameter search
backend/classification/optimization/optimizer.py
Other components
- Researcher
Connections 26
- Researcher → Streamlit app: opens
- Streamlit app → Dataset exploration: presents
- Streamlit app → Segmentation page: presents
- Streamlit app → Representation page: presents
- Streamlit app → Classification page: presents
- Dataset exploration → Dataset loader: uses
gui/pages/0_exploration.py - Segmentation page → Dataset loader: uses
- Representation page → Dataset loader: uses
gui/pages/2_representation.py - Classification page → Dataset loader: uses
gui/pages/3_classification.py - Representation page → Method controls: uses
gui/pages/2_representation.py - Classification page → Method controls: uses
gui/pages/3_classification.py - Dataset loader → Smart-home dataset: loads
gui/components/data_loader.py - Smart-home dataset → Events and segments: builds events
backend/dataset/methods/base.py - Method controls → Activity annotators: runs annotator
gui/components/shared.py - Classification page → Segmentation methods: runs segmenter
gui/pages/3_classification.py - Classification page → Hyperparameter search: configures search
gui/pages/3_classification.py - Classification cross-validation → Smart-home dataset: splits dataset
backend/classification/quality/cv.py - Classification cross-validation → Segmentation methods: fits and transforms
backend/classification/quality/cv.py - Classification cross-validation → Feature extractors: extracts features
backend/classification/quality/cv.py - Classification cross-validation → Activity annotators: labels segments
backend/classification/quality/cv.py - Classification cross-validation → Classifiers: fits and predicts
backend/classification/quality/cv.py - Classification cross-validation → Segmentation metrics: scores boundaries
backend/classification/quality/cv.py - Classification cross-validation → Classification metrics: scores labels
backend/classification/quality/cv.py - Hyperparameter search → Classification cross-validation: evaluates candidates
backend/classification/optimization/optimizer.py - Segmentation validation → Segmentation methods: validates methods
backend/segmentation/quality/cv.py - Representation page → Representation analysis: analyzes results
gui/pages/2_representation.py