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

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

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.