How daveebbelaar/tracking-barbell-exercises works
Main Components
• Data Sources – Represent the raw CSV files housed in the “PythonCode/data” directory (and subfolders like “form”). These are produced from the Mbientlab’s sensor device.
• Data Ingestion & Preprocessing – Include components from Chapter2 (“CreateDataset.py”) where the raw data is initially ingested and converted.
• Data Transformation & Cleaning – Cover Chapter3 scripts (e.g., “DataTransformation.py”, “ImputationMissingValues.py”, “KalmanFilters.py”, “OutlierDetection.py”). This stage involves cleaning the data, imputing missing values, filtering noise, and dealing with outliers.
• Feature Extraction (Abstraction) – Incorporate Chapter4 modules (“FrequencyAbstraction.py”, “TemporalAbstraction.py”, “TextAbstraction.py”) which extract different types of features from the sensor signals.
• Clustering & Distance Metrics – Add Chapter5 (e.g., “Clustering.py”, “DistanceMetrics.py”) that likely cluster similar exercises or patterns.
• Evaluation & Model Training – Represent Chapter7 (with “Evaluation.py”, “FeatureSelection.py”, “LearningAlgorithms.py”, “PrepareDatasetForLearning.py”) and Chapter8 (“LearningAlgorithmsTemporal.py” along with the dynsys subfolder code “Evaluator.py”, “Model.py”) where the machine learning models are built, selected, and evaluated.
• Visualization & Reporting – Indicate utility modules (e.g., “VisualizeDataset.py”, “ClassificationReport.py”) and example notebooks that output graphs and reports. These correspond to the “PythonCode/images” directory and notebook files which display the results.
Relationships and Interactions
• Then, indicate that the processed features feed into two branches: one for clustering analysis (Chapter5) and another for evaluation/model training (Chapter7 & Chapter8).
• Finally, highlight the role of “Visualization & Reporting” in consuming outputs from both the clustering and learning modules.
• Also, show that intermediate results are stored (see “PythonCode/intermediate_datafiles”) which may be reused or passed between stages.
Architectural Patterns and Design Principles
• Pipeline Architecture – Emphasize the clear sequential and modular steps (data ingestion, cleaning, feature extraction, and model training) that follow a pipeline pattern.
Relevant Technologies and Frameworks
• Python – The entire code base (scripts and notebooks) is in Python.
• Jupyter Notebooks – For exploratory analysis and demonstration of different data processing and learning steps.
• Machine Learning Libraries – Implicitly referenced by modules (e.g., clustering, learning algorithms) though specific libraries aren’t named. It’s likely that common libraries (numpy, pandas, scikit-learn, etc.) are used.
• Conda and Pip – Represent the deployment/runtime environment where dependencies are managed via “conda_requirements.txt” and “pip_requirements.txt”.