NORA: an Open-Access 3D Multimodal Dataset for the Oil & Gas Field

The NORA dataset is the first publicly available, multi-modal 3D dataset representing a full-scale industrial mock-up from the Oil & Gas (O&G) sector. Developed through a collaboration between SENAI Innovation Institute for Sensing Systems (ISI-SIM) and an O&G industry partner, NORA addresses a long-standing gap in accessible data for research and development in industrial environments—particularly where security and proprietary constraints have traditionally made data sharing nearly impossible. The dataset comprises: Non-Annotated Point Cloud (raw.e57): Captured using a FARO Focus S 150 terrestrial LiDAR scanner as part of a Design of Experiments (DoE)-driven acquisition strategy, this file contains the raw point cloud data from all 18 individual scan positions. The scans are preserved separately within the .e57 format, each enriched with scan pose metadata. The scanned structure is a full-scale (15 m × 7.5 m × 5 m) industrial mock-up representing a piping and instrumentation replica. This format supports detailed analysis of scan geometry, occlusion effects, and multi-view registration. Annotated Point Cloud (nora.ply): This consolidated point cloud merges the 18 registered scans into a unified model and includes per-point semantic and instance labels. Each point is annotated with both a class label (e.g., valve, flange, structural support, instrumentation) and a unique instance ID to distinguish between repeated components of the same class. The .ply format is compatible with common 3D processing and machine learning frameworks, making it suitable for tasks such as semantic segmentation, instance segmentation, object detection, and structural analysis. CAD Model (cad.obj): A clean, parametric model representing the design-phase geometry of the same structure, useful for comparing as-designed vs. as-built environments. P&ID (pnid_X_XXX.pdf): A piping and instrumentation diagram providing functional and logical mapping of the system. Drawn using standard symbol conventions in line with ISO 10628 and ISA 5.1 standards, it serves as a bridge between physical layout and operational logic. Panoramic Images (FARO_Scan_XXX.png): High-resolution, HDR color images captured at each of the 18 scan positions. Aligned with the 3D scans, these images support multimodal tasks such as component recognition via OCR, occlusion validation, and visual inspection. Acknowledgment: This research was carried out in association with the ongoing R&D project registered with ANP under No. 24520-9, entitled “3D APT: 3D Automated Perception and Tagging” (SENAI/Shell Brazil/ANP). The project is sponsored by Shell Brasil Petróleo Ltda. under the ANP R&D levy (“Compromisso de Investimentos com Pesquisa e Desenvolvimento”) and is also supported by financial resources from grant No. 016/2017, signed between the SENAI Innovation Institute for Sensing Systems and EMBRAPII.