Neural Network

TerraMind: Multimodal Foundation Models for Earth Observation Tasks

OnlineDiscovery - GeoAIWorkshop

  • Register

     

    * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

    • Date
    18 November 2025
    Timeframe
    14:00 - 16:00 CET Geneva
    Duration
    120 minutes
    Share this session

    TerraMind, co-developed by IBM and ESA’s Φ-lab, is the first generative, multimodal foundation model for Earth observation, designed to push the boundaries of geospatial intelligence. It introduces Thinking-in-Modalities (TiM) for seamless reasoning across diverse data sources and achieves state-of-the-art performance on community benchmarks. In this AI for Good seminar, we will explore the principles of foundation models and demonstrate how TerraMind works in practice. Using the TerraTorch toolkit, participants will generate multimodal outputs and fine-tune TerraMind for a real-world downstream application in an interactive, hands-on session. 

     

    Instructions for the workshop: 

    Participants only need a reliable internet connection for model downloads or to connect via Colab. If you have a local machine with a GPU (such as a recent MacBook) or access to a GPU cluster, you can prepare by cloning https://github.com/IBM/terramind and following the setup instructions. Alternatively, you can join the seminar using Colab without any prior setup; however, this option may be slightly less reliable and require some installations during the session. 

    • Share this session
    • Days
      Hours
      Min
      Sec
    Add to Calendar
    Discover more from programme stream
    Discovery - GeoAI
    TerraMind: Multimodal Foundation Models for Earth Observation Tasks

    TerraMind: Multimodal Foundation Models for Earth Observation Tasks

    18 November 2025 - 14:00 to 16:00

    Related stories

    Are you sure you want to remove this speaker?