Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
The dataset viewer is not available for this split.
Parquet error: Scan size limit exceeded: attempted to read 312488717 bytes, limit is 300000000 bytes Make sure that 1. the Parquet files contain a page index to enable random access without loading entire row groups2. otherwise use smaller row-group sizes when serializing the Parquet files
Error code:   TooBigContentError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

ORION WWF1 – Representative Sample Pack v1.0

1. Dataset Summary

ORION WWF1 is a high-fidelity industrial dataset designed for the next generation of industrial AI...

Visual Preview (Sample Clips)

Clip SRC_021 Clip SRC_029 Clip SRC_039
High-Intensity Anonymization Privacy-First Industrial Vision Industrial POV Perspective

2. Dataset Description

The dataset captures real-world operations within a Wooden Window Factory (WWF1). Unlike general-purpose video datasets, ORION focuses on the "Industrial POV," capturing complex interactions between workers, machinery, and raw materials.

  • Primary Goal: Facilitate the development of AI models that respect worker privacy without compromising technical accuracy.
  • Key Feature: Irreversible high-intensity anonymization certified for industrial compliance.

3. Dataset Contents

The sample pack includes:

  • 10 Video Clips: 1080p @ 30fps, 30 seconds each.
  • Anonymization Metadata: Stability logs and tracking fingerprints.
  • Full Taxonomy: Hierarchical classification of industrial assets.
  • Technical Schemas: Formal JSON/Markdown definitions for all metadata fields.

4. Dataset Structure

ORION_WW_PACK/
├── samples/              # Deliverables (MP4 + QA Logs)
├── manifest/             # TSV mapping of all assets
├── integrity/            # SHA-256 verification files
├── orion_wwf1_fields_schema.json  # Metadata specification
├── orion_wwf1_taxonomy.json       # Asset hierarchy
├── DATASHEET.md          # Technical Deep-Dive
└── LICENSE               # Usage Terms

5. Data Collection

Data was collected using calibrated industrial cameras (POV/Head-mounted) within a functioning wood-processing plant.

  • Environment: High-noise, variable lighting, industrial machinery.
  • Personnel: Certified operators performing routine production tasks.

6. Annotation Methodology

While this sample pack focuses on the video stream, the underlying ORION methodology employs:

  • Temporal Consistency: Annotations are tracked across frames with EMA-smoothed trajectories.
  • Privacy-First Masking: All human identifiers are masked using a dual-model approach (Face detection + Head silhouette scanning).

7. Quality Assurance (QA) Process

Our Zero False Negative policy ensures every human subject is protected:

  1. AI Predetection: Automated scan for all human presence.
  2. Human-in-the-loop (HITL): 100% manual review of all detection tracks.
  3. QA Certification: Automated verification of blur intensity (Kernel 99x99) for every frame.

8. Schema Description

The dataset utilizes a standardized schema (v1.1) to ensure interoperability with industrial MLOps pipelines. Key fields include:

  • clip_id: Unique identifier.
  • anonymization_verified: Boolean certification.
  • industrial_context: Metadata regarding the specific factory zone.

9. Taxonomy

The ORION Taxonomy covers:

  • EPI (PPE): Helmets, Gloves, Protection glasses.
  • Tools: Hand tools, Power tools.
  • Machinery: Saws, CNC machines, Assembly lines.

10. Dataset Figures (Full Dataset Roadmap)

Feature Sample Pack (This) Full Dataset (In Development)
Clip Count 10 Clips 622 clips (v1.x extended)
Total Duration 5 Minutes ~5.2 Hours
Anonymization 100% Certified 100% Certified
Screenshots N/A 9000+ High-Res with BBox/Seg
Physical AI N/A Action & Context Description
Advanced HAR N/A Human Activity Recognition

Physical AI, HAR, and EPI detection are currently in active development and are not included in this demo pack.

11. Use Cases

  • Safety Analysis: Detecting PPE and dangerous maneuvers.
  • Production Efficiency: Cycle time analysis in manufacturing.
  • Privacy Research: Benchmarking anonymization robustness.

12. Limitations

  • Volume: Being a sample pack, it is not intended for large-scale training.
  • Class Depth: Restricted to the wood-processing sector.

13. Licensing

Distributed under the ORION Sample Data License. Usage is permitted for internal R&D and professional evaluation. Redistribution is prohibited.

14. Citation

ORION Dataset: wooden_window_factory_01 (wwf1). ORION – Industrial AI Data Lab (2026).

15. About ORION

ORION – Industrial AI Data Lab is at the forefront of Securing Industrial Data and Empowering Intelligence. We bridge the gap between industrial privacy and AI performance.


Official Channels: orion-the-lab.com | LinkedIn

Downloads last month
90