How ciglenecki/eeg-driver-fatigue-detection works
This repository implements an EEG driver-fatigue research workflow: prepare CNT recordings, filter and transform signals into epoch-level features, clean and normalize a dataframe, train and compare classifiers, then rank significant electrodes. The README documents entropy combinations, participant-held-out SVM tuning, four classifier families, and model reporting; sampled sources show the dataframe, normalization, model configuration, and electrode-weight calculations. Several training and postprocessing files are indexed but not excerpted, so documented capabilities are included while uncertain internal call links are omitted.
Main entry points: postprocess_significant_electrodes.py Electrode Analysis · preprocess_create_df_from_signals.py Dataframe Builder · dataframes Feature Dataframes · model_summary.py Model Summary
Components 24
Signal Preparation
- Unpack Data
src/preprocess_unzip_data.py - Signal Processing
src/utils_signal.py - Signal Configuration
src/utils_env.py
Feature Data
- Feature Extraction
src/utils_feature_extraction.py - Dataframe Builder
src/preprocess_create_df_from_signals.py - Data Cleaning
src/preprocess_preprocess_df.py
Model Training
- Classifier Configuration
src/model.py - Entropy Search
src/train_entropies.py - SVM Tuning
src/train_svm_params.py - Model Training
src/train_models.py
Evaluation Analysis
- Model Summary
src/model_summary.py - Electrode Analysis
src/postprocess_significant_electrodes.py - All-Driver Weights
src/postprocess_significant_electrodes_all.py - Per-Driver Weights
src/postprocess_significant_electrodes_users.py
Shared Runtime
- Artifact Persistence
src/utils_file_saver.py - Workflow Utilities
src/utils_functions.py - Path Configuration
src/utils_paths.py - Plotting
src/utils_plot.py - Dataframe Statistics
src/utils_df_stats.py
Other components
- Researcher
- EEG Recordings
- Feature Dataframes
data/dataframes - Saved Models
models - Reports
reports
Connections 40
- Researcher → Unpack Data: prepares data
README.md - EEG Recordings → Unpack Data: provides recordings
README.md - Unpack Data → Dataframe Builder: supplies signals
- Signal Configuration → Dataframe Builder: configures processing
src/preprocess_create_df_from_signals.py - Dataframe Builder → Signal Processing: filters and crops
src/preprocess_create_df_from_signals.py - Dataframe Builder → Feature Extraction: extracts features
src/preprocess_create_df_from_signals.py - Dataframe Builder → Feature Dataframes: saves dataframes
src/preprocess_create_df_from_signals.py - Feature Dataframes → Data Cleaning: loads dataframe
src/preprocess_preprocess_df.py - Data Cleaning → Feature Dataframes: saves cleaned data
src/preprocess_preprocess_df.py - Feature Dataframes → Entropy Search: provides training data
src/train_entropies.py - Feature Dataframes → SVM Tuning: provides training data
src/train_svm_params.py - Feature Dataframes → Model Training: provides training data
src/train_models.py - Classifier Configuration → Entropy Search: configures SVM
src/train_entropies.py - Classifier Configuration → SVM Tuning: provides parameter ranges
src/train_svm_params.py - Classifier Configuration → Model Training: configures classifiers
src/train_models.py - Entropy Search → Reports: reports combinations
src/train_entropies.py - SVM Tuning → Reports: reports parameters
src/train_svm_params.py - Model Training → Saved Models: saves models
src/train_models.py - Model Training → Reports: reports performance
README.md - Saved Models → Model Summary: supplies models
src/model_summary.py - Model Summary → Reports: writes summaries
src/model_summary.py - Feature Dataframes → Electrode Analysis: provides dataframe
src/postprocess_significant_electrodes.py - Saved Models → Electrode Analysis: provides SVM
src/postprocess_significant_electrodes.py - Electrode Analysis → All-Driver Weights: dispatches all mode
src/postprocess_significant_electrodes.py - Electrode Analysis → Per-Driver Weights: dispatches driver mode
src/postprocess_significant_electrodes.py - Electrode Analysis → Reports: reports weights
src/postprocess_significant_electrodes.py - Dataframe Builder → Artifact Persistence: saves artifacts
src/preprocess_create_df_from_signals.py - Artifact Persistence → Workflow Utilities: formats metadata
src/utils_file_saver.py - Dataframe Builder → Workflow Utilities: uses helpers
src/preprocess_create_df_from_signals.py - Dataframe Builder → Path Configuration: uses data paths
src/preprocess_create_df_from_signals.py - Data Cleaning → Artifact Persistence: saves cleaned data
src/preprocess_preprocess_df.py - Data Cleaning → Path Configuration: uses dataframe path
src/preprocess_preprocess_df.py - Model Summary → Artifact Persistence: loads and saves
src/model_summary.py - Model Summary → Workflow Utilities: uses helpers
src/model_summary.py - Model Summary → Path Configuration: uses report paths
src/model_summary.py - Electrode Analysis → Artifact Persistence: loads model
src/postprocess_significant_electrodes.py - Electrode Analysis → Workflow Utilities: uses helpers
src/postprocess_significant_electrodes.py - Electrode Analysis → Path Configuration: uses report paths
src/postprocess_significant_electrodes.py - Researcher → Model Training: requests evaluation
README.md - Researcher → Electrode Analysis: requests analysis
README.md