CLIP Embeddings of 562,842 Library of Congress Historical Map Sheets (Sanborn maps, atlases, city views)
- Provider
- Library of Congress Computing Cultural Heritage (Huang and Lee 2025); embeddings by Mahowald (2024)
- License
- CC-BY-4.0 (source)
- Format
- JSON (in sampled files)
- Stored at
- Amazon S3, s3://us-east-2.opendata.source.coop/l4cch/loc-maps-clip (us-east-2)
- Access
- Free and public. No account with the data provider is needed.
- DOI
- 10.5281/zenodo.11538437
- Category
- Machine Learning & AI
- Source
- https://doi.org/10.1017/chr.2025.10017
About this dataset
One CLIP ViT-B/32 image embedding for each of 562,842 scanned map sheets in the Library of Congress Geography and Map Division, from 56,525 catalog items. 78% are Sanborn fire insurance maps of American towns; the rest are city views, atlases, charts and other maps, nine in ten dated 1862 to 1955. data/embeddings/embeddings.pt is a PyTorch float32 tensor of 562,842 x 512 L2-normalized vectors; item_ids.pt lists each sheet's IIIF identifier in vector order; metadata.json gives each sheet's IIIF image service, thumbnail and full-size request URLs, its loc.gov record and its title and date. Images are not included and can be fetched at any size from tile.loc.gov. It is the index behind the Digital Collections Explorer for text-to-image and image-to-image search of historical cartography. Load with torch.load and search with cosine similarity. Licensed CC BY 4.0 (the Zenodo embeddings record); cite Huang and Lee (2025), Computational Humanities Research 1:e14, and Mahowald (2024), doi:10.5281/zenodo.11538437.
How to access this dataset in Python
import boto3
from botocore import UNSIGNED
from botocore.client import Config
s3 = boto3.client('s3', region_name='us-east-2', config=Config(signature_version=UNSIGNED))
s3.list_objects_v2(Bucket='us-east-2.opendata.source.coop', Prefix='l4cch/loc-maps-clip/', Delimiter='/')Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
File types in sampled files: JSON
Files
Example file: l4cch/loc-maps-clip/data/embeddings/embeddings.pt
Analyze this dataset without downloading it
On Scigantic, CLIP Embeddings of 562,842 Library of Congress Historical Map Sheets (Sanborn maps, atlases, city views) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.