A Scalable AI Training Platform for Remote Sensing Data

These slides I presented at AGILE 2023 confernece in Delft. They show our work on a scalable platform for AI training based on the example of remote sensing data.
Workflow management systems like Steep are well suited for such use cases. They can flexibly provide computational power and can be easily extended to cover the additional requirements of AI training (like reproducibility).
https://doi.org/10.5194/agile-giss-4-53-2023

Transcript

1. A Scalable AI Training Platform for Remote Sensing Data — June 15th, 2023 Hendrik M. Würz, Kevin Kocon, Barbara Pedretscher, Eva Klien, Eva Eggeling
2. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 2 Forest Shape File from Bayerische Vermessungsverwaltung CC BY 4.0 Satellite image produced from ESA remote sensing data. Copernicus Sentinel 2, Captured 2022-06-15, Retrieved from SciHub 2023-05-05
3. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 3 Forest Shape File from Bayerische Vermessungsverwaltung CC BY 4.0 Satellite image produced from ESA remote sensing data. Copernicus Sentinel 2, Captured 2022-06-15, Retrieved from SciHub 2023-05-05
4. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 4 Forest Shape File from Bayerische Vermessungsverwaltung CC BY 4.0 Satellite image produced from ESA remote sensing data. Copernicus Sentinel 2, Captured 2022-06-15, Retrieved from SciHub 2023-05-05 On-Premises Flexibility Reproducibility & Versioning Language agnostic
5. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 5 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
6. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 6 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
7. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 7 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
8. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 8 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
9. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 9 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
10. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 10 On-Premises Flexibility Reproducibility & Versioning Language agnostic Architecture
11. Thank you for listening —
12. Contact — Hendrik M. Würz GEO Department hendrik.martin.wuerz@igd.fraunhofer.de Fraunhofer IGD Fraunhoferstraße 5 64283 Darmstadt, Germany www.igd.fraunhofer.de
13. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 13 Forest Shape File from Bayerische Vermessungsverwaltung CC BY 4.0 https://geodaten.bayern.de/opengeodata/OpenDataDetail.html?pn=tatsaechlichenutzung (Retrieved 2023-05-05) Satellite image produced from ESA remote sensing data. Copernicus Sentinel 2, Captured 2022-06-15, Retrieved from SciHub 2023-05-05 Gear Icon by Design Circle - Flaticon, Faticon license https://www.flaticon.com/de/kostenloses-icon/zahnrad_7424705 (Retrieved 2023-05-05) AI Icon by Freepik - Flaticon, Flaticon license https://www.flaticon.com/de/kostenloses-icon/ki_5207500 (Retrieved 2023-05-05) Checkmark by Google (modified) - Flaticon, CC 3.0 BY, https://www.flaticon.com/de/kostenloses-icon/prufzeichen_60516 (Retrieved 2023-05-15) Image Sources
14. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 14 Training Workflow
15. Public IGD-Folienvorlage-de.potx, Version 3.4, 17.06.2022 © Fraunhofer IGD 2023-06-15 Page 15 Usage Workflow