Input
Input single or multiple AOIs and Dates and choose the model. S2DR software fetches the corresponding Sentinel-2 imagery automatically from ESA Copernicus repository.
Products
Self-serve super-resolution for Sentinel-2
Sentinel-2 Deep Resolution (S2DR) transforms freely available Sentinel-2 imagery from 10 m to 1 m using deep learning — reconstructing real spatial detail rather than interpolating it, across all 10 spectral bands. It comes as two models, S2DR3 and S2DR4, accessible via REST API or Python SDK and billed from a single credit balance.
Super-resolution models
Two models, one product
Both S2DR3 and S2DR4 models enhance Sentinel-2 imagery to 1 m across all 10 bands. They differ in how they handle cloud and haze.
S2DR3 — the sharpest results. Best for clear skies.
S2DR4 — maximum usability. Reliable even in cloudy conditions.
Sentinel-2 · 10 m
S2DR4 · 1 m
Sentinel-2 true-colour imagery of a residential district in Los Angeles at 10 m, beside the GammaEarth S2DR4 output at 1 m, where the street grid, the freeway corridor and single rooftops become legible.
How it works
Processing pipeline
AI processing
The selected model (S2DR3 or S2DR4) super-resolves all 10 Sentinel-2 bands from their native 10 m and 20 m to a uniform 1 m in a single pass. Bands are processed jointly, reconstructing sub-pixel detail while preserving inter-band radiometric consistency.
Output
16-bit multiband GeoTIFFs, generated locally (SDK) or fetched from our API bucket.
Software interfaces
Two access paths for technical and operational users
Integrate through the REST API or Python SDK, with both S2DR models billed from one credit balance.
REST API
Submit jobs and retrieve output programmatically. Full endpoint documentation available.
View API docsPython SDK
Choose the Python package to run super-resolution and write outputs in your own environment. Requires Linux x86-64, Python 3.12, CUDA GPU.
View SDK docsEvaluating for integration?
Qualified users can request hands-on access to the API and SDK to test before committing.