Deliverable 3.1 Annotated Datasets

Deliverable 3.1 of the DISCOVER project (HORIZON-CL4-2023-TWIN-TRANSITION 01-11-101129909) defines the data governance methodology and technical infrastructure underpinning the DISCOVER project’s approach to generating and managing annotated datasets for BIM 4.0. The central goal is to support the creation of semantically enriched, interoperable data products derived from multimodal sources (images, point clouds, sensors) to enable advanced applications such as predictive maintenance and asset lifecycle optimization. The governance strategy is built around ETL (Extract, Transform, Load) pipelines organized in three phases: • Metadata Generation & Image Annotation: Led by Tecnalia, this phase extracts structured metadata from visual datasets (e.g., RGB, XRF) provided by partners (TUD, UPC) in WP1 and WP2, preparing them for downstream integration. Key outputs include JSON-formatted image metadata with location, time, and device data. • IFC Generation from Point Clouds: Led by DTT, raw point clouds from Lidar and photogrammetry are transformed into geometric models in the IFC format. A robust protocol ensures data quality, spatial referencing, and adequate density for Level of Development (LOD) 200 BIM models. • IFC Enrichment: Metadata and contextual image information are spatially and temporally matched with IFC elements to produce a fully enriched BIM 4.0 model. This integration supports semantic querying and interoperability across use cases. A modular reference architecture supports deployment flexibility. It includes a cloud based stack using FastAPI, Apache Airflow, MongoDB, PostgreSQL, and FileZilla, and is virtualized via OpenStack. Each use case operates a tailored pipeline instance with shared governance principles and templates. The infrastructure ensures scalability, traceability, and compliance with FAIR and GDPR requirements. While real datasets are still in development, this deliverable provides a detailed methodological and technical blueprint to ensure data consistency, quality, and reuse across the project lifecycle.