05 Sep

An Auto-ML framework for EO

  • Intelligent Software Applications
  • Technical R&D Consulting

Earth Observation (EO) covers everything from drones to satellite imagery, and its applications  have a huge impact on everyday life: predicting weather, creating maps, tracking container ships, monitoring agriculture and so much more. These applications require the processing of an enormous amount of data (for instance, the European Sentinel-2 satellite covers the whole globe in every 5 days, generating 10 TB of daily data) . This amount of data can only be processed with automated methods.

Deep learning, the most successful branch of artificial intelligence (AI), relies on both a high amount of carefully curated data, as well as  specialised models. This leads to a problem: expensive experts are required both for data collection and for designing appropriate deep learning models. If a small company would like to utilise AI for customised mapping, that would require significant expertise and cost.

The goal of the ATELIER-EO (Automated machine learning framework tailored to Earth Observation) project is to develop an automated artificial intelligence solution for EO. More effective learning schemes will reduce the amount of required data, while the automated machine learning will reduce the necessary deep learning expertise and the domain specific solutions will improve performance of the models.

As an example, the mapping of forest roads is currently very challenging: roads are small, and while a human knows they are connected, the tree cover often hides parts of the road. This makes roads very challenging for a model to find. ATELIER-EO will be able to automatically handle this connectivity, as well as incorporating terrain models and mitigate other issues  like clouds in the images. The only input required will be training images where example roads are present. 

A far more complex challenge is the detection of tree species in forests  based on the shapes of the tree. This is a task far beyond the capability of most non-experts. Most people might recognize that the red and yellow regions have different species, but only a few would be able to tell that the red polygon has spruce and the yellow contains pine forest.

In conclusion, ATELIER-EO will be able to automatically create customised maps from satellite or drone images with limited data and minimal human interaction. This will broaden the domains where AI can be applied,  reducing the cost of customised products.