Last_min_clutch_force is a semi-supervised signals intelligence (SIGINT) pipeline designed to identify known friendly signals and detect unknown/hostile emitters using raw RF IQ waveforms.
In a contested environment, operators often have labeled data for their own forces but no examples of what adversary signals look like. This system solves the problem by treating signal classification as an anomaly detection task.
- HDF5 Ingestion: Efficiently reads raw 256-element IQ signal vectors at scale.
- Feature Engineering: Extracts statistical markers from raw waveforms, including amplitude variance, energy distribution, and signal envelope characteristics.
- ML Anomaly Detection: Uses an Isolation Forest model trained on friendly signal distributions to flag out-of-distribution (hostile/unknown) emitters with confidence scores.
- Tactical Dashboard: An interactive Common Operating Picture (COP) map that geolocates and color-codes emitters in real-time.
- Streamlit - Front-end dashboard framework.
- Scikit-Learn - Machine learning engine (Isolation Forest).
- h5py - HDF5 dataset processing.
- Folium - Interactive geospatial visualization.
- NumPy & Pandas - Signal processing and data management.
- Clone the repository.
- Install dependencies:
pip install streamlit pandas numpy h5py scikit-learn folium streamlit-folium