BDAPPV: Aerial Images of Rooftop Photovoltaic Installations in France and Belgium

BDAPPV is a dataset of aerial images showing rooftop photovoltaic installations. It includes segmentation masks and installation metadata, sourced from two aerial imagery providers, Google and IGN. The dataset was created by Gabriel Kasmi and published in Scientific Data in 2023.

Use Cases

• Train semantic segmentation models based on provided segmentation masks for rooftop PV installations.
• Benchmark image classification models based on aerial imagery of PV installations.
• Evaluate model robustness to distribution shift based on imagery from two different sources (Google and IGN).
• Analyze the geographic distribution of PV installations based on the metadata covering France and Belgium.

Strengths

• Includes segmentation masks, enabling pixel-level analysis.
• Contains installation metadata, providing context beyond the imagery.
• Sourced from two distinct providers (Google and IGN), facilitating domain shift studies.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the coverage of France and Belgium.

Provenance

Source

Aerial imagery from Google and IGN.

Collection Method

Collection and annotation described in the associated paper by Kasmi et al.

Freshness

Last updated 2026-06-05 16:32:27; freshness should be verified.

Geography

France and Belgium.