Comparison between Sentinel-2 RGB imagery (left in each pair) and PCA projections of the learned feature embeddings (right) for twelve representative locations.

New preprint for the #DeepFeatures project:
Why choose between spatial detail and temporal fidelity in #EO data?

We developed a #ContextAware #multimodalAI framework that unifies satellite data ( #Sentinel1 & #Sentinel2) into a single, high-resolution view of land surface properties.

#AI4Science

Poster time! You can find us #DeepFeatures at T17 and T18.

#LPS25 #LPS2025 #Poster #ESA

The #DeepFeatures team is at the #LPS25 with two poster in Zone T on Thursday. Come by & find out more 🤓

With Julia Peters, @dmlmont.bsky.social @cmosig.bsky.social @gdkrmr.bsky.social Martin Reinhardt, Gunnar Brandt, Konstantin Ntokas, Teja Kattenborn, Daria Svidzinska, @miguelmahecha.bsky.social

#LPS25 here I come 🚆!
Another magnificent train journey to #Vienna despite the summer heat. Excited to present our project #DeepFeatures to you.

Funded by #ESA #AI4Science

#LPS2025 #DeepLearning #ML #EO #multispectral #Sentinel2 #SpectralIndeces #DimensionReduction #Spatiotemporal #Phenology

#DeepFeatures News 📣
We are excited that our PhD student Julia Peters presented first project results at the #ESA - #NASA International Workshop on #AI #FoundationModel for #EO.