How kee-rti/safetyappmlmodelprototype works
This repository is a Streamlit ML prototyping platform for safety-risk classification from synthetic multi-sensor data. Its main workflow runs from scenario generation and preprocessing through configurable fusion/model training, evaluation, and result interpretation or ONNX export. Experiment history and model versioning are also surfaced in the application. The README documents the product stages; sampled source confirms key imports, training/evaluation calls, and ONNX validation/benchmarking. Some module internals were not sampled, so their interactions are shown only where documented.
Main entry points: app.py Streamlit workspace · data_loader.py Dataset loader · safetyAppDataset_1755171350672.py Scenario generator · models.py Neural architectures
Components 12
Data workflow
- Scenario generator
attached_assets/safetyAppDataset_1755171350672.py - Dataset loader
src/data_loader.py - Feature preprocessing
src/preprocessing.py
Modeling
- Fusion strategies
src/fusion_techniques.py - Neural architectures
src/models.py
Analysis and deployment
- Training and evaluation
src/evaluation.py - Model interpretation
src/interpretability.py - ONNX export
src/onnx_export.py
Application and sensor support
- Streamlit workspace
app.py - Experiment tracking
app.py - Sensor specifications
utils/sensor_specs.py
Other components
- Platform user
Connections 15
- Platform user → Streamlit workspace: uses
README.md - Streamlit workspace → Scenario generator: generates scenarios
README.md - Scenario generator → Dataset loader: provides samples
src/data_loader.py - Dataset loader → Scenario generator: imports generator
src/data_loader.py - Streamlit workspace → Dataset loader: loads data
README.md - Streamlit workspace → Feature preprocessing: preprocesses data
README.md - Streamlit workspace → Fusion strategies: selects fusion
README.md - Streamlit workspace → Neural architectures: selects architecture
README.md - Streamlit workspace → Training and evaluation: trains and evaluates
app.py - Streamlit workspace → Model interpretation: analyzes results
README.md - Streamlit workspace → ONNX export: exports and benchmarks
app.py - Streamlit workspace → Experiment tracking: records runs
app.py - Streamlit workspace → Sensor specifications: uses sensor definitions
app.py - Dataset loader → Sensor specifications: uses specifications
- Neural architectures → Fusion strategies: applies fusion module