PCFootprint is a large-scale, open-source benchmark for vectorized building footprint extraction from Airborne Laser Scanning (ALS) point clouds. It contains 227,264 annotated building instances across 33,000 standardized tiles, covering a geographic extent of 540.67 km². The dataset was created by Haoyuan-Shen and sourced from the Estonian Land and Spatial Development Board, with a last recorded update on 2026-03-13.
Use Cases
- Train and benchmark models for vectorized building footprint extraction based on ALS point cloud data.
- Develop algorithms for urban planning and land-use analysis based on the annotated building instances.
- Research methods for processing large-scale, standardized geospatial tiles for feature detection.
Strengths
- Large scale with 227,264 annotated building instances.
- Extensive geographic coverage of 540.67 km² across 33,000 tiles.
- Specifically engineered as an open-source benchmark for a defined task.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- Estonian Land and Spatial Development Board
- Collection Method
- Airborne Laser Scanning (ALS) point clouds, processed and annotated.
- Time Range
- null
- Freshness
- Last updated 2026-03-13 14:44:57; freshness should be verified.
- Geography
- Estonia, covering 540.67 km²