High-Resolution Post-Disaster Remote Sensing Imagery and Building Footprint YOLO-Segment D

High-resolution optical satellite imagery and pre-disaster building footprint annotations for two disaster events in 2025. The dataset contains one large post-disaster image per event, with building footprints converted to YOLO-Segment format, produced by Zhimin Wu and released in 2026. It includes imagery from the Beijing-3/N satellite for the Dingri earthquake and WorldView-3 for the California wildfire.

Use Cases

• Train building footprint segmentation models based on high-resolution post-disaster satellite imagery.
• Benchmark geographic registration and alignment techniques based on pre- and post-disaster data.
• Develop disaster impact assessment workflows based on building location and footprint information.
• Test YOLO-Segment model performance on real-world, high-resolution remote sensing data.

Strengths

• Spatial resolution is explicitly stated (0.75 m for Dingri, 0.31 m for California wildfire).
• Dataset size is 2.8 GB, indicating substantial image data.
• No missing data exists, and spatial errors were minimized through geographic registration.
• Data is provided in standard formats (.tif for imagery, .txt for YOLO-Segment annotations).

Limitations

• Building damage-level labels are not included, limiting direct use for damage assessment.
• Original building vector (shp) files were not released, restricting alternative analyses.
• Row count is unknown, which may limit suitability assessment for large-scale training.

Provenance

Source

figshare, author Zhimin Wu.

Collection Method

Produced through image organization, preprocessing of pre-disaster building vector data, coordinate system unification, geographic registration, format conversion, and visual inspection.

Time Range

Imagery acquired in January 2025.

Freshness

Last updated 2026-05-25 08:28:19; freshness should be verified.

Geography

Dingri, Tibet, China and California, United States.