Skip to content

SCynergy 2026

SCynergy 2026 - GeoAI Workshop — Overview

Author: Eun-Kyeong Kim (eun-kyeong.kim@lxp.lu), LuxProvide S.A

What is this workshop about?

This workshop introduces how multimodal geospatial data can be used together with a foundation model (TerraMind) to perform flood mapping. This tutorial is given in the context of the Scynergy 2026 event and EPICURE.

You will learn how satellite data from different sources (Sentinel-1, Sentinel-2, and DEM) can be:

  • discovered
  • aligned
  • transformed
  • and used for AI-based prediction

Learning Objectives

By the end of this workshop, you will:

  • Understand what GeoAI and geospatial foundation models (GeoFMs) are
  • Learn why multimodal data is important for Earth observation tasks
  • Understand how geospatial data must be aligned and prepared before AI can be applied
  • Run a TerraMind flood model on real satellite data
  • Interpret model predictions and their limitations

Workshop Structure

Part 1 — Concepts (20 minutes) - Presentation Slides (.pdf)

  • GeoAI and multimodal Earth observation
  • TerraMind foundation model
  • Geospatial data concepts (raster, reprojection, alignment)
  • End-to-end workflow overview

Part 2 — Hands-on (60 minutes) - Learning Materials (notebooks)

You will work through three notebooks and one optional notebook:

  1. Data Acquisition: Search and select satellite data using STAC
  2. Data Packaging: Align and transform data into model-ready format
  3. Model Inference: Run TerraMind to detect flooded areas
  4. (Optional) Fine-Tuning: IBM Tutorial - Fine-tune existing TerraMind models with Sen1Floods11 dataset

Key Idea

👉 AI models are only as good as the data they receive.

A major focus of this workshop is understanding how geospatial preprocessing enables multimodal AI.

Platform

This workshop runs on:

  • MeluXina GPU cluster
  • Open OnDemand JupyterLab

No prior HPC experience is required.

Who is this for?

This workshop is designed for:

  • AI / ML practitioners interested in geospatial data
  • Remote sensing / GIS experts exploring AI methods
  • Researchers working with Earth observation data

What you will build

By the end, you will have:

  • Executed a complete GeoAI pipeline
  • Understood how TerraMind works in practice
  • Gained a reusable workflow for other EO applications

Takeaway

👉 From satellite catalog → aligned data → AI prediction

Resources

EPICURE LuxProvide