license: apache-2.0
tags:
- open-science
- hugging science
- dataset-index
- scientific-ml
- ml-for-science
- metadata
- physics
- biology
- chemistry
- astronomy
- materials-science
- csv
Datasets Index
This dataset is a curated collection of scientific datasets from across the web, compiled to support open-source machine learning in science. Many of these datasets are difficult to access, scattered across domain-specific websites, or stored in inconvenient formats. The goal of this index is to make them more discoverable for researchers, engineers, and contributors to the HuggingFace4Science community.
Dataset Overview
Rows: 52 datasets/urls
Format: CSV file
Fields:
id – Unique identifier
created_at – Timestamp of entry creation
dataset_name – Human-readable dataset name
dataset_url – Source link
description – Short summary of dataset contents
approx_size – Estimated dataset size
size_unit – Unit of size (e.g., MB, GB, TB)
field – Scientific domain (e.g., physics, materials science, biology)
user – Contributor who submitted the dataset
Use Cases
Discovery: Find datasets relevant to ML in physics, chemistry, materials science, and beyond.
Preprocessing: Many listed datasets are in raw formats (NetCDF, FITS, domain-specific archives). This CSV provides a starting point for building standardized loaders.
Community Contribution: Expand the list by submitting new entries, cleaning existing datasets, or publishing derived versions to Hugging Face Datasets Hub.
How to Contribute
Suggest new datasets via the Hugging Science community.
Contribute loaders, enrichment scripts, or preprocessed dataset cards.
Help validate metadata (sizes, formats, licenses).
Disclaimer
This index does not host the actual datasets. It only provides links and metadata. Users are responsible for checking each dataset’s original license, distribution restrictions, and usage terms before downloading or using the data.
License
The index file is released under Apache 2.0. Individual datasets retain their original licenses. Please check the linked sources before use.