--- title: Maya4 emoji: 📊 colorFrom: indigo colorTo: indigo sdk: static pinned: true license: apache-2.0 thumbnail: >- https://cdn-uploads.huggingface.co/production/uploads/68a43a86eea45496edb28ba6/hJyITwdmrofMtEtX8qFmW.png ---

Maya4 Logo

Maya4

Multi-level intermediate SAR representations from Sentinel-1 Stripmap acquisitions

Level-0 to Level-1 · Zarr-native · Cloud-accessible · 2 TB

--- ## Overview **Maya4** is a curated SAR data resource designed to expose the full progression of **Sentinel-1 Stripmap acquisitions** from **raw radar echoes** to **fully focused Level-1 imagery**. Unlike conventional datasets that provide only final products, Maya4 preserves and organizes the **intermediate signal representations** generated across the SAR processing chain. This makes the dataset particularly suitable for: - SAR signal processing research - physics-aware machine learning - self-supervised pre-training - representation learning across processing levels - algorithm benchmarking and reconstruction studies The name *Maya4* draws from the concept of **Māyā**: the idea that reality is revealed through successive layers. In the same way, SAR imagery emerges through a sequence of transformations from raw measurements to interpretable image products. --- ## Why Maya4

Multi-level access

Provides consistent access to multiple SAR processing stages rather than only the final image product.

Research-oriented structure

Designed for analysis of information flow, model pre-training, and development of custom SAR pipelines.

Cloud-native delivery

Distributed in Zarr format for scalable storage, streaming, and computation.
--- ## Dataset Access | Dataset | Access | Mission / Mode | Format | Size | |---------|--------|----------------|--------|------| | **Maya4** | [Open bucket](https://huggingface.co/buckets/ESA-philab/Maya4) | Sentinel-1 Stripmap | Zarr | 2 TB | --- ## Processing Chain A defining feature of Maya4 is its **sharded multi-level organization**, which preserves the major intermediate states of the SAR focusing pipeline.

Maya4 intermediate SAR representations

| Processing Level | Abbrev. | Description | Technical Value | |------------------|---------|-------------|-----------------| | **Raw** | `raw` | Unprocessed radar echoes as acquired by Sentinel-1 | Enables custom end-to-end SAR processing and low-level signal analysis | | **Range Compressed** | `rc` | Echoes compressed in the range dimension using matched filtering | Improves signal-to-noise ratio and resolves scatterers along range | | **Range Cell Migration Corrected** | `rcmc` | Echoes after compensation of range migration effects | Preserves geometric consistency and prepares the signal for azimuth focusing | | **Azimuth Compressed** | `ac` | Fully focused SAR image in slant-range geometry | Corresponds to the interpretable focused SAR image product | --- ## Technical Positioning Maya4 is intended to support work at the intersection of: - SAR signal processing - remote sensing foundation models - self-supervised and masked modeling approaches - physics-guided representation learning - inverse problems and reconstruction - benchmarking of processing-aware architectures Because the dataset exposes multiple internal stages of SAR formation, it enables experiments that are not possible with image-only repositories. --- ## Key Characteristics | Attribute | Value | |-----------|-------| | **Mission** | Copernicus Sentinel-1 | | **Acquisition Mode** | Stripmap | | **Processing Coverage** | Level-0 to Level-1 intermediates | | **Primary Distribution Format** | Zarr | | **Access Paradigm** | Cloud-native bucket access | | **Primary Target Users** | SAR researchers, ML practitioners, remote sensing scientists | --- ## Acknowledgements Maya4 is based on data from the **Copernicus Sentinel-1 mission** of the **European Space Agency (ESA)**. Dataset curation and organization are maintained by the **Maya4 organization**.