# Earth Index Embeddings V7: SSL4EO-S12 DINO ViT-S/16 Sentinel-2 chip embeddings, global, 2024 and 2025-2026 (Earth Genome, Source Cooperative)

## Facts

- Provider: Earth Genome
- License: CC-BY-4.0 (https://source.coop/earthgenome/earthindex-embeddings-v7)
- Format: Parquet (in sampled files)
- Stored at: Amazon S3, s3://us-west-2.opendata.source.coop/earthgenome/earthindex-embeddings-v7 (us-west-2)
- Access: public, readable without an account
- Category: Earth Observation & Geospatial
- Source: https://earthindex.ai
- Page: https://scigantic.com/archive/abd780da-43c9-592b-8cf2-2b0ba82354f4
- Keywords: Earth Observation & Geospatial, satellite imagery, Parquet, AWS Open Data

## About

Chip-level vector embeddings of yearly cloud-free Sentinel-2 mosaics for the whole land surface between about 60.5 degrees south and 70.3 degrees north, produced by Earth Genome for Earth Index, a search engine that finds places that look like a place you already know. Each embedding is a 384-dimensional float32 vector from the self-supervised SSL4EO-S12 DINO ViT-S/16 model (loaded through TorchGeo), computed for one Major TOM grid cell of about 320 x 320 m read from the 10 m mosaic. Two yearly periods sit side by side: 2025-08-16 to 2026-08-16 (17,849 files, from the Earth Index Imagery V7 mosaic) and 2024-01-01 to 2025-01-01 (17,847 files, from the earlier V4 mosaics). Files are GeoParquet, one per MGRS tile and period (for example 31TDF_2025-08-16_2026-08-16.parquet, a few hundred MB each), each carrying a STAC item with the embeddings (emb:) extension in its footer. Use them for similarity search, clustering, change detection or as features for land-cover classifiers. Vectors are not comparable with the earlier SoftCon-based earthindexembeddings product. Read with DuckDB, geopandas or pyarrow. Licensed CC BY 4.0 by Earth Genome.

## Load the data

```python
import boto3
from botocore import UNSIGNED
from botocore.client import Config
s3 = boto3.client('s3', region_name='us-west-2', config=Config(signature_version=UNSIGNED))
s3.list_objects_v2(Bucket='us-west-2.opendata.source.coop', Prefix='earthgenome/earthindex-embeddings-v7/', Delimiter='/')
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: Parquet

Example file: `earthgenome/earthindex-embeddings-v7/01GDM_2024-01-01_2025-01-01.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/abd780da-43c9-592b-8cf2-2b0ba82354f4
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/abd780da-43c9-592b-8cf2-2b0ba82354f4
- Over MCP: `get_schema_card` with id `abd780da-43c9-592b-8cf2-2b0ba82354f4` at https://api.scigantic.com/mcp

## Related topics

- [Earth Observation & Geospatial datasets](https://scigantic.com/topics/earth-observation-and-geospatial)
- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Satellite imagery datasets](https://scigantic.com/topics/satellite-imagery)
