PaCaBa: Parking Cars Barcelona Satellite Image Dataset

A WorldView-3 stereo satellite image dataset with labeled parking cars covering roughly 5 km² across four connected regions in Barcelona. The dataset includes raw geotiff images with polygon annotations and 540x540 image patches with rotated bounding boxes for 7303 parking cars and over 12,000 general cars. It was created by Sebastian Zambanini of TU Wien.

Use Cases

• Train a parking car detector model based on the provided 540x540 image patches with rotated bounding box annotations.
• Test a general car detection model using the stereo image patches containing both moving and static vehicles.
• Develop geospatial analysis tools for urban planning based on the raw geotiff images with polygon annotations covering street areas.

Strengths

• Contains annotations for 7303 parking cars and over 12,000 general cars across stereo images.
• Covers approximately 5 km² of urban area across four distinct regions in Barcelona.
• Provides data in multiple formats suitable for different tasks: raw geotiffs and pre-processed 540x540 patches.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Data may reflect geographic bias inherent to its single-city (Barcelona) scope.

Provenance

Source

Sebastian Zambanini, TU Wien

Collection Method

WorldView-3 stereo satellite imagery, manually annotated.

Geography

Four connected regions in Barcelona, Spain.