Building Footprints in Accra, Ghana with 40,786 Structures

1,330 high-resolution image tiles and corresponding vector labels provide building footprint data for Accra, Ghana. This dataset was created by MLHUB and published via NASA EarthData, with a last update in January 2023. It is a Tier 1 dataset from the RAMP project, indicating it has been thoroughly reviewed and improved.

Use Cases

• Train building segmentation models based on high-resolution satellite imagery and vector label pairs.
• Benchmark urban density analysis based on the labeled building footprint data.
• Develop geospatial machine learning models for applications in developing regions, based on the dataset's focus on Accra.

Strengths

• Contains 1,330 image-label tile pairs for model training.
• Includes labels for 40,786 individual buildings.
• Described as a Tier 1 dataset, indicating a thorough review and improvement process.
• Imagery has a high spatial resolution of 30 cm per pixel.

Limitations

• Column-level documentation and sample data are unavailable, requiring field semantics to be inferred after download.
• The dataset's size in gigabytes and specific file formats beyond .tif and .geojson are unknown.
• Last updated 2023-01-01 00:00:00; freshness should be verified.

Provenance

Source

MLHUB, via NASA EarthData.

Collection Method

Originally sourced from the Open Cities AI Challenge Dataset, with drone imagery resampled to 30 cm and labels improved.

Freshness

2023-01-01 00:00:00

Geography

Accra, Ghana.