dataset · Global

VHR-10

Hugging Face dataset satellite-image-deep-learning/VHR-10

Claim this listing

Represent VHR-10? Send a claim for review so we can discuss listing updates. Claims do not automatically grant ownership or verification.

Data offered docs-verified: object detection bounding boxes and instance masks for 10 classes: airplanes, ships, storage tanks, baseball diamonds, tennis courts, basketball courts, ground track fields, harbors, bridges, vehicles · VHR optical remote sensing images (715 colour, 85 pansharpened colour infrared)
Coverage llm-inferred: Global
Resolution docs-verified: 0.5 to 2 m (Google Earth colour images); 0.08 m (pansharpened colour infrared images from Vaihingen)
Formats docs-verified: Parquet (auto-converted by the Hugging Face Dataset Viewer)
Access docs-verified: Free/open; MIT licence; no authentication listed in ENTRY metadata

Use cases llm-inferred: Benchmarking object detection of bridges, harbours and storage tanks in VHR imagery for site appraisal screening · Prototyping automated vehicle or sports-facility detection to support land-use audits · Training or evaluating instance segmentation models for GIS feature extraction workflows
Audience llm-inferred: GIS analysts · remote sensing specialists · data scientists / ML engineers · planners

Constraints llm-inferred: Mixed sources and resolutions (0.5 to 2 m Google Earth versus 0.08 m Vaihingen), which complicates a single model or workflow · Ten object classes only; no drainage, flood, or water-specific features · No UK-specific coverage is documented · Small dataset: the description cites 800 images while the viewer shows 809 rows, so counts should be checked · Hugging Face copy is a mirror of a GitHub repository, so it may fall out of step with the source

Verified 2026-10-15 via HF API: exists; 556 downloads/30d, 3 likes.

Metadata observed 14/10/2026 · research sweep: manual-2026-10-15 (Hugging Face datasets)