Simplifying SAR for AI Development

ICEcube: an open source python library for SAR-based machine learning

Build your ICECUBE external_link

About The AI4SAR Project

Rapid-revisit Synthetic Aperture Radar (SAR) satellites in Low Earth Orbit (LEO) are promising with their observational persistence, low cost, and breakthrough engineeringHowever, the complexity of the backscattered SAR signal is quick to overwhelm both novice and expert data scientists.


At ICEYE Analytics, we are keen on building AI/ML applications for heavy-duty image processing and scalable analytics. The European Space Agency (ESA) Φ-lab presented us with an opportunity to lower the entry barrier to SAR-based ML applications. Due to the combination of our internal objectives and this fantastic ESA opportunity, we’ve begun to develop an open source Python library to easily enable ML exploration of ICEYE SAR images.

The ICEcube array handling library is our first step to lower the SAR barrier to entry. This library abstracts, yet persists, some of the metadata handling and optimizes array manipulation and memory usage through xarray data structures.



  • SAR Processing
  • SAR Processing

    SAR-ML pipelines support quick iterations that enable us to identify and correct SAR processing artifacts. ML applications can be deployed easily to a cloud processing environment and can scale asynchronously to handle large quantities of data.
  • Time Series Analysis
  • Time Series Analysis

    Time-series stacks of coregistered SAR images help us appreciate the true weight of the challenge at hand. SAR is the only EO technology that can consistently and with high precision enable the quantification of change over time. High-temporal resolution and persistence under cloud cover create opportunities to study the dynamics (manmade or natural) of the Earth regardless of season. Unsanctioned deforestation can no longer be hidden in the rainy season.
  • Data Handeling
  • Data Handling

    SAR preprocessing quickly overwhelms with its inherent complexity, unique processing chain, and stringent requirements, such as coregistration, to make sense of the data. ICEcube abstracts away this burden.




Muhammad Irfan Ali

Machine Learning Engineer
Irfan on LinkedIn


Andrea Radius

SAR Expert
Andrea on LinkedIn


Tapio Friberg

Machine Learning Engineer 



Shay Strong

VP, Analytics
Shay on LinkedIn


Bartosz Postulka

Sr. Technology Manager


Arnaud Dupeyrat

Data Scientist
Arnaud on LinkedIn


Ibrahim El Merehbi

Remote Sensing Engineer
Ibrahim on LinkedIn


Leticia Mendonca

Technical Writer
Leticia on LinkedIn




Blog: Applying Machine Learning To Rapid Revisit SAR

The complexity of SAR data is often daunting and discourages even experts from integrating it into their Machine Learning workflows. We at ICEYE Analytics want to change that.

Read more
Build Your OWN ICEcube

Extract the essence of SAR without the burden of heavy-duty preprocessing

Build It Now