gabrielireland/local-embeddings-2022

Local Embeddings Dataset

Multi-temporal satellite imagery dataset for phenology embedding training.

Dataset Description

This dataset contains multi-spectral satellite tiles across 6 months (April-September 2022) with 16 bands per tile.

Dataset Structure

local_embeddings/
├── alphaearth_embeddings_tiles_202204/  (263 tiles)
├── alphaearth_embeddings_tiles_202205/  (263 tiles)
├── alphaearth_embeddings_tiles_202206/  (263 tiles)
├── alphaearth_embeddings_tiles_202207/  (263 tiles)
├── alphaearth_embeddings_tiles_202208/  (263 tiles)
└── alphaearth_embeddings_tiles_202209/  (263 tiles)

Data Format

• Format: GeoTIFF (.tif)
• Bands per file: 16
  • Bands 1-15: Feature channels (spectral/radar data)
  • Band 16: Ground truth labels
• Tile naming: ae_rXXX_cXXX_s2_YYYY-MM-DD_s1_YYYY-MM-DD.tif
• Spatial tiles: 260 tiles present across ALL 6 months

Statistics

• Total files: 1578
• Dataset size: 7.31 GB
• Months covered: April - September 2022
• Time series: 6 months per spatial location

Usage

from pathlib import Path
import rasterio
from huggingface_hub import snapshot_download

# Download dataset
dataset_path = snapshot_download(
    repo_id="gabrielireland/local-embeddings-2022",
    repo_type="dataset"
)

# Load a tile
tile_path = Path(dataset_path) / "alphaearth_embeddings_tiles_202204" / "ae_r000_c000_s2_2022-04-24_s1_2022-04-17.tif"
with rasterio.open(tile_path) as src:
    data = src.read()  # Shape: (16, H, W)
    features = data[:15]  # Bands 1-15
    labels = data[15]     # Band 16

Citation

If you use this dataset, please cite:

@dataset{local_embeddings_2022,
  title={Local Embeddings Multi-Temporal Dataset},
  year={2022},
  url={https://huggingface.co/datasets/gabrielireland/local-embeddings-2022}
}

License

This dataset is released under CC-BY-4.0 license.

---

Uploaded: Training_UNet_Alpha_Earth_Pipeline