MIT Lincoln Laboratory LADI

Use in your agent

Load LADI v2 test labels into pandas and Parquet. Checked against the provider's documentation on Nov 1, 2026.

Access

• No AWS account or key. S3 listing and object downloads are public.
• This example downloads only the 268 KB v2a test-label CSV, not the image archive.

Request

curl -O 'https://ladi.s3.us-west-2.amazonaws.com/Labels/v2/ladi_v2a_labels_test.csv'

Format

CSV with source image URL, local path and binary labels for bridges, buildings, damage thresholds, debris, flooding, roads, trees and water.

Load into pandas and Parquet

Needs Python 3 with requests, pandas and pyarrow.

"""LADI v2a test labels -> pandas -> Parquet. No key."""
import io
import requests
import pandas as pd

URL = "https://ladi.s3.us-west-2.amazonaws.com/Labels/v2/ladi_v2a_labels_test.csv"
r = requests.get(URL, timeout=60)
r.raise_for_status()
df = pd.read_csv(io.BytesIO(r.content))
df.to_parquet("ladi_v2a_test_labels.parquet", index=False)
print(len(df), "labeled images", len(df.columns), "columns", "first columns", ", ".join(df.columns[:5]))

Run by us on 2026-11-01 (python 3.13.5 | pandas 3.1.0 | pyarrow 26.0.0 | requests 2.34.2). Result: 1163 labeled images 14 columns first columns url, local_path, bridges_any, buildings_any, buildings_affected_or_greater

Limits

• Rate limit: Not stated.
• The v2a test-label CSV is about 268 KB; image archives range into gigabytes.
• Updates are described only as periodic.

Quirks

• The CSV stores image locations rather than embedding image bytes; downloading images is a separate and much larger step.
• The v2 test split comes from 2023 and intentionally differs from the training and validation event distribution.
• v1 and v2 use different label sets and annotation processes; do not merge their target columns without mapping them.

Terms and attribution

• AWS lists the dataset under CC BY 4.0.
• Use the registry's stated access-date citation when crediting the AWS dataset.

Sources

• https://registry.opendata.aws/ladi/: “Creative Commons Attribution 4.0 International (CC BY 4.0)”
• https://registry.opendata.aws/ladi/: “AWS CLI Access (No AWS account required)”
• https://github.com/LADI-Dataset/ladi-overview: “The v2 dataset consists of approximately 10k images”
• https://github.com/LADI-Dataset/ladi-overview: “The images are geotagged (in their EXIF metadata).”