ModularGeoFM is a personal, early-stage project exploring how foundation models can be applied to geospatial and Earth observation (EO) data.
The main goal is to experiment with modular components and data pipelines that support training, fine-tuning, or adapting foundation models for various EO tasks.
The project is in an early exploratory phase, with emphasis on:
- Reusable, modular data pipelines for satellite and remote sensing data
- Experimenting with components of foundation models adapted to EO tasks
- Investigating scalable pretraining, fine-tuning, and inference strategies
Feel free to browse, fork, or reach out if you're exploring similar ideas.
⚠️ This is not a production-ready framework. It’s an open-ended space for prototyping and learning.