RFInject/README.md

---
name: RFInject
authors:
  - Roberto Del Prete
  - Nermine Hendy
license: APACHE-2.0
source: https://cdn-avatars.huggingface.co/v1/production/uploads/6538dbdcaca9fa108fd50595/2nTO6KAHh2UvbaJTa0jCs.png
thumbnail: https://philab.esa.int/
---

# RFInject – Synthetic RFI Injection for Sentinel-1 SAR Data

## Overview

RFInject is a research-grade Earth Observation dataset and methodology for controlled synthetic Radio Frequency Interference (RFI) injection into clean Sentinel-1 Synthetic Aperture Radar (SAR) data.

The dataset is designed to enable reproducible benchmarking of RFI detection and mitigation algorithms, addressing a long-standing gap in the SAR community caused by the lack of standardized and controllable interference datasets.

RFInject preserves the physical and statistical properties of real Sentinel-1 acquisitions while enabling full parametric control over injected interference characteristics.

## Motivation

Radio Frequency Interference is a major source of performance degradation in modern SAR missions. Sentinel-1 data is particularly affected, yet most existing studies rely on ad-hoc or irreproducible contamination scenarios.

RFInject enables:
- Repeatable experimental setups
- Controlled and parameterized interference scenarios
- Algorithm-agnostic benchmarking across methods and sensors

## Dataset Structure

```
/
├── RFInject/                 # Sentinel-1 data with injected synthetic RFI
│   ├── product_001.zarr
│   ├── product_002.zarr
│   ├── product_003.zarr
│               └── burst_0
│               └── burst_1
│               └── burst_2
│                   └── zarr.json  (the product metadata)
│                   └── echo (the clean burst)
│                   └── rfi  (the rfi to add to burst)
│                       └── zarr.json  (the RFI metadata)
└── README.md
```


## Data Characteristics

| Property            | Description |
|---------------------|-------------|
| Platform            | Sentinel-1 |
| Sensor              | C-band SAR |
| Data Level          | L0 |
| Interference Type   | Synthetic RFI (parametric) |
| File Format         | zarr / analysis-ready |


## Methodology

Synthetic RFI is injected by superimposing parameterized interference signals onto clean Sentinel-1 radar echoes.

The approach ensures:
- Spectral and temporal realism
- Preservation of system characteristics
- Full reproducibility through metadata-controlled parameters

## Ingestion with EOTDL

### CLI

```bash
eotdl datasets get RFInject
```

## Intended Use

This dataset is intended for:
- RFI detection and mitigation research
- Machine learning model training and evaluation
- Algorithm benchmarking
- Reproducible SAR processing experiments

Users are responsible for validating suitability for operational or safety-critical applications.

## Citation

```
@misc{rfinject_2025,
	author       = { RFInject },
	title        = { v1 (Revision 23853e6) },
	year         = 2025,
	url          = { https://huggingface.co/datasets/RFInject/v1 },
	doi          = { 10.57967/hf/7227 },
	publisher    = { Hugging Face }
}
```

## License

This dataset is released under the APACHE-2.0 license.

## Acknowledgements

Developed within the ESA Φ-lab research ecosystem and related collaborations.