The Open Geospatial Consortium (OGC) is excited to announce that the OGC Membership has approved the OGC Training Data Markup Language for Artificial Intelligence (TrainingDML-AI) Part 1: Conceptual Model for adoption as an official OGC Standard. The Standard defines the conceptual model for standardized geospatial training data for Machine Learning.
Training data plays a fundamental role in Earth Observation (EO) Artificial Intelligence Machine Learning (AI/ML) applications, especially Deep Learning (DL). It is used to train, validate, and test AI/ML models. Understanding the source and applicability of training data allows for better understanding of the results of AI/ML operations.
To maximize the interoperability and re-usability of geospatial training data, the TrainingDML-AI Standard defines a model and encodings consistent with the OGC Standards baseline to exchange and retrieve the training data via the Web. Part 1 of the Standard contains the Conceptual Model, as well as example JSON encodings. Future Parts of the Standard will cover other encodings.
Additionally, the Standard provides detailed metadata for formalizing the information model of training data. This includes but is not limited to the following aspects:
- How the training data is prepared, such as provenance and quality;
- How to specify different metadata used for different ML tasks;
- How to differentiate the high-level training data information model and extended information models specific to various ML applications;
- How to describe the version, license, and training data size;
- How to introduce external classification schemes and flexible means for representing ground-truth labeling.
OGC Members interested in staying up to date on future progress of this standard, or contributing to its development, are encouraged to join the Training Data Markup Language for AI Standards Working Group via the OGC Portal. Non-OGC members who would like to know more about participating in this SWG are encouraged to contact the OGC Standards Program.
As with any OGC standard, the open OGC Training Data Markup Language for Artificial Intelligence (TrainingDML-AI) Part 1: Conceptual Model Standard is free to download and implement.