Custom-built EO Monitoring

Organisations working with Sentinel-2 imagery often find 10 m pixels too coarse for the features they need to see, and lack in-house models to turn the imagery into decisions. Semablu addresses this with two linked offerings: an AI super-resolution product that sharpens free Copernicus Sentinel-2 imagery, and bespoke AI models built for a customer's specific Earth Observation monitoring task.

The imagery product is analytics-grade, orthorectified surface-reflectance RGB+NIR data derived from Sentinel-2 MSI L2A at 2.5 m spatial resolution, with coverage equivalent to Sentinel-2 availability. It is bought through Semablu's own console: draw an area of interest, pay per km² from prepaid credits (€0.25/km² pay-per-scene at the time of review, credits from €10, 100 free credits on sign-up), and receive GeoTIFFs, with the archive discoverable through a free STAC API that works with pystac-client and standard STAC-aware GIS tooling. Semablu publishes a benchmark of its model against ESA's OpenSR (LDSR-S2) on the Sen2NAIP dataset (PSNR 36.09 vs 35.43 dB, SSIM 0.870 vs 0.851) and shows the gain carrying through to derived products such as NDVI.

Alongside the imagery, Semablu develops custom models on Earth Observation data for monitoring applications, drawing on funded R&D in the downstream EO sector: ESA-supported projects on soil-erosion monitoring (SEMS-2D) and Posidonia oceanica seagrass mapping (PosidoniaMapper) with super-resolved Sentinel-2 data, and the Xjenza Malta-funded PI-EOS project on super-resolution for EO monitoring.

Key capabilities:

• Super-resolved Sentinel-2 imagery: 2.5 m orthorectified RGB+NIR surface reflectance from Sentinel-2 MSI L2A, delivered as GeoTIFF within minutes of ordering.
• Self-serve access: web console with AOI drawing and per-km² credit pricing, plus a free STAC API for programmatic discovery.
• Land cover mapping: custom models that produce land cover maps to support policy development, urban planning, and forest and agricultural monitoring.
• Change detection: models that automatically pinpoint changes between repeat satellite acquisitions over urban or rural areas.
• Object detection: custom detectors for buildings, vessels, airplanes, PV panels and other targets in remote sensing imagery.