Our Technology

The core elements for the AI4EO* revolution

*Artificial Intelligence for Earth Observation

Our focus is an end to end learning approach: in goes data, out goes results, obtaining a time reduction in data preparation and processing of several orders of magnitude. Our time performance improvement is translated in affordable services.

Data fusion

By combining multiple data sources (optical, radar; satellite, UAV; high and low resolution) and leveraging temporal information we can take out the most from the available data at any time without sacrificing performance.

Continuous learning

Our automated labelling tool enables fast generation of new data as it is available in order to re-train and improve our models on a daily basis.

Self-supervised learning

We pre-train our models with EO data, instead of using models pre-trained on natural imagery. This improves the data efficency of the models, obtaining very good results with smaller labelling efforts.

Custom architectures

We design and develop Neural Networks adapted to EO data peculiarities, working with multi-modal data and giving predictions for different tasks while maximizing the model capacity and parameter sharing.

DL models ready to use

Classification, segmentation and detection models to measure densities, quantify targets and evaluate changes. All accessible through SPAI.