Major-TOM: Expandable Datasets for Earth Observation

Major-TOM functions as an extensible framework specifically designed to standardize the curation, sharing, and combination of large-scale Earth Observation (EO) datasets. It addresses the current fragmented landscape of EO data by providing a shared system that allows users to efficiently access and integrate multiple datasets. This initiative aims to minimize the duplication of effort often involved in compiling data for data-hungry deep learning models, fostering greater interoperability within the EO community.

The platform hosts a substantial collection of official image and embedding datasets. These resources encompass multi-spectral imagery from Sentinel-2, Synthetic Aperture Radar (SAR) data from Sentinel-1, and Digital Surface Models (DSM) derived from Copernicus DEM. Many of these datasets are global in scope and are provided in formats specifically optimized for machine learning applications, including extensive sets of pre-computed embeddings.

Major-TOM actively encourages further contributions to its project, cultivating a community-driven approach to building open and compatible EO resources. A key element of its design is an underlying geographical indexing system, which enables the seamless integration and expansion of these vast data collections. This framework is vital for advancing research and development in both Earth observation and artificial intelligence, offering a robust foundation for future innovations.