Building Footprints in Cox's Bazar, Bangladesh with 35 cm Satellite Imagery

A chipped training dataset contains 2,375 image tiles and corresponding vector labels for 26,875 individual buildings. It was created by MLHUB and last updated in January 2023. The dataset includes high-resolution satellite imagery in .tif format and building footprint labels in .geojson format.

Use Cases

• Train building footprint segmentation models based on high-resolution 35 cm satellite imagery.
• Validate remote sensing models for peri-urban and rural areas based on the described land-use keywords.
• Develop localized geospatial AI models based on the dataset's use in testing a 'ramp' model for Cox's Bazar.
• Conduct land-use analysis based on the presence of labeled agricultural, peri-urban, and rural building footprints.

Strengths

• Includes 2,375 pre-aligned image and label tile pairs.
• Labels 26,875 individual building footprints.
• Uses high-resolution satellite imagery at 35 cm per pixel.
• Is designated as a 'ramp Tier 1' dataset, indicating it has been thoroughly reviewed and improved.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2023-01-01 00:00:00; freshness should be verified.

Provenance

Source

MLHUB via NASA Earthdata

Collection Method

Satellite imagery sourced from Maxar ODP (10400100546E3700).

Freshness

Last updated 2023-01-01 00:00:00

Geography

Cox's Bazar, Bangladesh