Demo ↗
MP4 · 76.4 MB
SI² — SuperIntelligence for Science
BLUEPRINT
Multimodal retrieval for complex engineering drawings and documents.
Research question
How can researchers find relevant engineering drawings when legacy metadata is incomplete?
Research approach
Layout-aware retrieval combines drawing-region detection, vision-language OCR, normalized identifiers, and lexical and dense search.
Research outputs & focus
- Searchable engineering records
- Structured metadata for legacy archives
- Retrieved drawings and documents
Artifacts
People behind the project
Contributors.
Related publications
The research behind the tool.
BLUEPRINT Rebuilding a Legacy: Multimodal Retrieval for Complex Engineering Drawings and Documents
Citation
Ethan Seefried et al. (2027). BLUEPRINT Rebuilding a Legacy: Multimodal Retrieval for Complex Engineering Drawings and Documents. WACV 2027.