BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//compute-paris-2026//speaker//AFZ7CQ
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20251125T000000
TZNAME:CET
TZOFFSETFROM:+0100
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20261025T030000
RDATE:20271031T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Navigating the Vanguard: A Practical Guide to Selecting Geospatial
  Foundation Models - H. Gijs J. van den Dool
DTSTART;TZID=Europe/Paris:20261125T144000
DTEND;TZID=Europe/Paris:20261125T151000
DTSTAMP:20260930T113518Z
UID:pretalx-compute-paris-2026-R3FUZL@pretalx.com
DESCRIPTION:The recent explosion of Earth Observation Foundation Models (E
 O-FMs)\, such as AlphaEarth\, TerraMind\, TESSERA\, and Copernicus-FM\, ha
 s introduced unprecedented capabilities in GeoAI. However\, their adoption
  presents considerable challenges\, including high computational demands\,
  diverse multimodal data requirements\, and substantial obstacles to model
  interpretation.\n\nTo address this "model fatigue\," this talk offers a p
 ractical guide to facilitate transparent and efficient model selection for
  data practitioners. We will survey the current landscape of GeoAI models
 \, demonstrate methods for evaluating predictive performance using accessi
 ble open-source tools\, and present alternative architectures that enable 
 scalable (parallel) processing on standard cloud GPUs\, eliminating the ne
 ed for high-performance computing clusters to test model accuracy on small
 er data domains and feature combinations. FMs are often presented as the s
 ingle truth\, but findings from combining feature sets from multiple EO-FM
 s at the prediction level suggest that ensemble approaches outperform any 
 single model.\n\nFinally\, we will address essential semantic and temporal
  considerations to ensure attendees understand the hidden uncertainties in
  their geospatial embeddings before making operational decisions\, and why
  the responsible use of AI-derived data requires that automated outputs ne
 ver be treated as ground truth. Foundation Models are designed for specifi
 c purposes\, and despite being termed "Foundation Models\," there is no si
 ngle model that fits all use cases — project domains are often highly sp
 ecific\, and closing that gap may require targeted model modification to g
 enerate custom embeddings that address what standard models miss.
LOCATION:Auditorium
URL:https://pretalx.com/compute-paris-2026/talk/R3FUZL/
END:VEVENT
END:VCALENDAR