Ecosystem Approach: AI in X3D and X3D in AI
AI Support for Taxonomies, Vocabularies, and Metadata Integration in the X3D Framework
Web3D Consortium's Special Interest Group on AI with X3D aims to explore, develop, and promote the interation of artificial intelligence technologies within the X3D framework. If you are interested, please subscribe to the AI mailing list and join the discussions.
Artificial Intelligence offers powerful new capabilities for enhancing the X3D ecosystem, particularly in the areas of metadata enrichment, vocabulary alignment, and semantic interoperability. By integrating curated vocabularies directly into authoring tools such as X3D-Edit, and by training AI models on consistent, domain‑specific terminology, the Web3D community can significantly improve the quality, discoverability, and longevity of 3D content.
Training AI models on curated terminology, and maintaining high-quality repositories, where 3D content is not only visually rich but also semantically meaningful, discoverable, and interoperable. This strategic alignment of AI and X3D strengthens the foundation for next-generation applications in cultural heritage, scientific visualization, engineering, education, and beyond.
1. Identifying and Listing Taxonomies and Vocabularies of Interest
A foundational step in semantic enrichment is the identification of authoritative vocabularies relevant to 3D graphics, cultural heritage, geospatial data, medical visualization, engineering, and other X3D application domains.
Key Activities
- Survey existing domain ontologies, thesauri, and controlled vocabularies.
- Identify authoritative sources such as Getty AAT, CIDOC-CRM, Dublin Core, schema.org, and domain‑specific standards.
- Map vocabulary terms to X3D metadata structures.
- Maintain a curated list of vocabularies for community use.
Benefits
- Ensures consistent terminology across X3D models.
- Supports interoperability with cultural heritage, museum, and scientific data ecosystems.
- Enables AI systems to reason over structured, semantically rich 3D content.
2. Integrating Vocabulary Terms into X3D-Edit
X3D-Edit can serve as a powerful bridge between domain vocabularies and practical authoring workflows. By embedding curated terms directly into the tool, authors can easily insert standardized metadata values into their models.
Capabilities
- Auto-suggest metadata values from controlled vocabularies.
- Provide dropdown lists or search interfaces for vocabulary terms.
- Validate metadata entries against known taxonomies.
Impact
- Reduces authoring friction and improves metadata accuracy.
- Ensures consistent use of terms across projects and institutions.
- Strengthens long-term preservation and semantic clarity of X3D content.
3. AI Training with Queriable, Repeatable Terms of Reference
Once vocabularies are curated and integrated, they become a powerful foundation for AI training. Large Language Models (LLMs) can be trained or fine‑tuned on these structured terms to support intelligent querying, metadata generation, and semantic reasoning.
AI Capabilities Enabled
- Suggesting metadata based on scene content or user prompts.
- Automatically generating semantic annotations for 3D models.
- Answering queries using consistent terminology.
- Supporting cross-domain reasoning (e.g., cultural heritage + geospatial).
Advantages
- Ensures AI outputs remain aligned with community standards.
- Improves reliability and repeatability of metadata suggestions.
- Enables advanced search and discovery across X3D repositories.
4. Curated Training Repositories
To support high-quality AI behavior, the Web3D community can maintain curated repositories of:
- X3D models with rich metadata
- Vocabulary-aligned annotations
- Domain-specific examples
- Best-practice templates
Repository Functions
- Provide training data for AI models.
- Serve as reference examples for authors.
- Support validation and benchmarking of metadata workflows.
Long-Term Value
- Strengthens the semantic foundation of the X3D ecosystem.
- Ensures AI systems evolve alongside community standards.
- Encourages consistent, high-quality metadata practices.
Web3D Member Projects, Research and Community Resources
- X3D Object creation with LLMs
- X3D Examples and model in HuggingFace - for training!
- Derived from X3D Examples Archives (3800 models): valid X3D with metadata, provided in multiple functionally equivalent file encodings (XML, ClassicVRML, JSON, X3D ResourcesTurtle) and programming languages (Java, Python)
- Virginia Tech ARC - all open hugging face model hosted
- X3D Examples and model in HuggingFace - for training!
- Research and Education
- Synthetic data training
- Spatial composition and relationships
- 3D browsing assistants
- Increasing Web3D Accessibility with Audio Captioning | Proceedings of the 28th International ACM Conference on 3D Web Technology
- Summarizing big, complex data with user context for training LLMs and agents
- Prompt Engineering for X3D Object Creation with LLMs | Proceedings of the 29th International ACM Conference on 3D Web Technology
- 2025 Web3D Tutorial Update! Links to slides available, we need to record video
- Community Resources
- Proposals coordinated by Web3D Consortium - sponsored
- We have enough diverse assets that one or more proposals appear to be possible.
- Partnered strategies are powerful, funding helps more people engage in worthy work.
- Web3D 2026 - Workshop on 3D content generation using AI
- Web3D 2026 - AI Contest - Sponsored
- Using Consortium curated benchmark data for training AI models
- Excellent exposure, publicity and credit for all participants (especially winners) that helps them continue to advance their work
- Proposals coordinated by Web3D Consortium - sponsored
TODO
- Use Model Context Protocal (MCP)
