Thinking Earth Hackathon - Harnessing Copernicus Foundation Models to Decode Earth from Space

Vision-Language Models for EO: connect imagery and text to enhance EO data interpretation

Thinking Earth & EU Agency for the Space Program

Date: 29-30 Sept, 2025
Hosted at: BiDS 2025 (Big Data from Space), University of Latvia, Rīga, LV-1004 Latvia
Track Followed: EO Vision-Language Models (Track 3)
Organizer: ThinkingEarth (Horizon Europe Project)
Website: Devpost – ThinkingEarth Hackathon
Sponsers: Thinking Earth || EUSPA || ESA || EUMETSAT || JRC, European Commision || Institute for Environmental Solutions, Latvia || National Technical Universty of Athens || National Observatory of Athens || TU Munich || NVIDIA || Evenflow
Collaborator: Athanasios Trantas & Udit Asopa

Key Highlights:

• EO Vision-Language Models Exploration: Built and fine-tuned multimodal deep learning workflows combining satellite imagery and text for Earth observation understanding
• Natural Language Interaction with EO Data: Experimented with vision-language models (VLMs) for image captioning, visual question answering, and geospatial data retrieval
• AI for Earth Interpretation: Leveraged Copernicus datasets and foundation models to enhance explainability and accessibility of Earth observation data
• Interdisciplinary Collaboration: Worked alongside participants from AI, remote sensing, and atmospheric sciences to bridge language and visual modalities
• Innovation in EO Accessibility: Contributed toward future-facing research on using VLMs to make geospatial analytics more intuitive and interactive for scientists and policymakers

Beyond Dead Wood Hackathon - Data and AI in search of forest biodiversity

Ultrahack

Date: 9 March, 2024
Location: Maria01, Lapinlahdenkatu 16, Helsinki
Website: Offcial Website of Event
Sponsers: Business Finland || SITRA || Location Innovation Hub || GEOFORUM || Metsäteho Oy || Aalto University Startup Center || Vaisala || NLS || Arbonaut
Collaborator: Udit Asopa, Monika Sharma, Gabriel Chibuye & Vesa Haaparanta

Key Highlights:

• Finnish Forest Innovation: Tackled real-world forestry challenges using data science and geospatial insights
• NLS Point Cloud Utilization: Leveraged high-resolution open LiDAR data to analyze forest structure and deadwood distribution
• ML-Powered Forestry Solutions: Teams applied machine learning techniques to classify forest elements and detect biodiversity indicators
• Entrepreneurial Focus: Encouraged actionable ideas with commercial potential to benefit forest health and sustainability
• Collaborative Hackathon Environment: Brought together interdisciplinary teams combining programming, environmental science, and business