A field guide to 190 industrial datasets

What it does

This is a curated awesome-list tracking 190 public industrial datasets spanning manufacturing, oil and gas, energy, and chemical processing. Each entry is tagged with a standardized Version 2 metadata schema covering domain, asset type, data modality, task, annotation style, access method, size, and license. The project also publishes an HTML version for easier browsing than the raw markdown table.

The interesting bit

Industrial data is notoriously hard to find outside proprietary silos, and when it does surface it is scattered across Kaggle competitions, university repositories, and Zenodo archives with no consistent labeling. The maintainer has done the tedious work of normalizing all of that into a single filterable schema, turning a research chore into a lookup task.

Key highlights

• 190 datasets cataloged across sectors like aerospace, chemical, water utilities, and additive manufacturing
• Standardized metadata fields including modality (time series, image, tabular), task (anomaly detection, predictive maintenance, RUL estimation), and access type
• HTML site generated for navigation beyond the raw GitHub markdown
• Version 1 structure preserved at a fixed commit for backward compatibility
• Many entries include real production and lab testbed data, not just synthetic benchmarks

Caveats

• A significant number of entries show “Information not available” for size, year, or license, so the metadata coverage is incomplete
• Access varies widely: some datasets require Kaggle logins, competition terms, or official portal registrations
• The repository itself is a catalog with markdown summaries; it does not host the actual data files

Verdict

Worth bookmarking if you are building industrial ML benchmarks or need real-world time-series, image, or tabular data from factory floors and field equipment. Skip it if you are looking for a unified download API or guaranteed clean, ready-to-train pipelines.