OGC Requests Public Feedback on TrainingDML-AI Standard for Geospatial AI Data

The Open Geospatial Consortium (OGC) has announced a public comment period for its candidate Training Data Markup Language for Artificial Intelligence (TrainingDML-AI) Part 1: Conceptual Model Standard, a new initiative aimed at standardizing geospatial training data used in Artificial Intelligence (AI) and Machine Learning (ML) workflows. The proposed standard defines models and encodings that support consistent sharing, discovery, and reuse of training datasets that involve location or time components.

Training datasets play a critical role in Earth Observation (EO), deep learning, and spatial analytics applications. Ensuring that these datasets are well-documented, interoperable, and accessible helps improve the reliability, transparency, and reproducibility of AI/ML results. The TrainingDML-AI standard seeks to address these needs by providing a common framework for describing how geospatial training data is created, structured, and applied.

Key Obectives of the TrainingDML-AI Standard

The candidate standard introduces a conceptual model designed to support the exchange and retrieval of geospatial training data through web-based systems while maintaining compatibility with existing OGC standards. It also establishes a structured metadata framework to ensure that datasets are properly documented and usable across different machine learning environments.

Key aspects addressed by the standard include:

• Documentation of training data provenance, preparation methods, and quality
• Metadata structures tailored to different machine learning tasks
• Representation of dataset versions, licensing information, and data volume
• Integration of external classification systems and labeling approaches
• Support for consistent encoding and interoperability across platforms

Advancing Interoperability for Geospatial AI

Standardizing training data formats is essential for improving collaboration across organizations and research communities. By aligning training datasets with shared encoding rules and metadata structures, the TrainingDML-AI initiative enables easier data exchange, reduces preprocessing requirements, and supports the development of scalable geospatial AI workflows.

The proposed framework also helps organizations better evaluate model performance by clearly documenting dataset characteristics, which strengthens trust in AI-driven geospatial analyses and decision-support systems.

Public Comment and Participation

OGC invites stakeholders from industry, government, academia, and research organizations to review the candidate TrainingDML-AI standard and submit feedback during the public comment period. Community participation plays a vital role in refining the specification and ensuring it meets the needs of the rapidly expanding geospatial AI ecosystem.

Organizations interested in contributing to the ongoing development of the standard are encouraged to join the OGC Training Data Markup Language for AI Standards Working Group and participate in collaborative discussions that shape future geospatial interoperability frameworks.