How arya-02a/rfi_detection_pipeline works
This is a research-prototype Streamlit application for inspecting GMRT radio-observation FITS files and detecting, classifying, and visualizing RFI. Its main workflow loads visibility data, converts it to a time–frequency power waterfall, creates a robust-statistics mask, builds labeled patches, trains a classifier, and displays results. A separate FITS explorer provides file metadata and signal summaries. The README documents additional dashboard views; source excerpts are partial, so uncertain internal wiring is omitted.
Main entry points: main.py Pipeline Dashboard · analysis.py FITS Explorer · dataset_builder.py Patch Dataset Builder · model_training.py Random Forest
Components 18
Dashboard
- Pipeline Dashboard
main.py - Pipeline Visualizations
main.py - Inspection Visualizations
main.py
Signal Processing
- Power Waterfall
main.py - Robust RFI Mask
main.py - Standalone Detector
detection.py
Patch Classification
- Patch Dataset Builder
dataset_builder.py - Patch Split
dataset_builder.py - Feature Extraction
model_training.py - Heuristic Classes
model_training.py - Random Forest
model_training.py - Overlay Map
model_training.py - Saved Model
models - Generated Dataset
dataset_ml
FITS Inspection
- FITS Explorer
analysis.py - FITS Summaries
analysis.py
Other components
- Researcher
- GMRT FITS
Connections 22
- Researcher → Pipeline Dashboard: uses
- GMRT FITS → Pipeline Dashboard: provides data
- Pipeline Dashboard → Power Waterfall: loads and converts
- Pipeline Dashboard → Robust RFI Mask: detects RFI
- Power Waterfall → Patch Dataset Builder: provides power
- Robust RFI Mask → Patch Dataset Builder: provides mask
- Patch Dataset Builder → Patch Split: splits patches
- Patch Split → Generated Dataset: saves arrays
- Pipeline Dashboard → Feature Extraction: requests features
- Pipeline Dashboard → Heuristic Classes: requests labels
- Pipeline Dashboard → Random Forest: trains model
- Random Forest → Saved Model: saves model
- Pipeline Dashboard → Overlay Map: builds overlay
- Pipeline Dashboard → Pipeline Visualizations: renders results
- Researcher → Standalone Detector: uses
- GMRT FITS → Standalone Detector: loads and detects
- Researcher → FITS Explorer: requests inspection
- GMRT FITS → FITS Explorer: reads structure
- FITS Explorer → FITS Summaries: computes summaries
- Pipeline Dashboard → FITS Explorer: runs analysis
- Pipeline Dashboard → Inspection Visualizations: renders analysis
- FITS Summaries → Inspection Visualizations: supplies results