Diverse Weather DroneVehicle: UAV Object Detection Benchmark

Diverse Weather DroneVehicle is a UAV-based object detection benchmark derived from the VisDrone/DroneVehicle dataset. It provides COCO-style annotations organized into four weather and illumination domains: day, dark, extreme_dark, and foggy. The dataset was created by author mingbohong and last updated on HuggingFace in May 2026.

Use Cases

• Training object detection models for robust performance across weather conditions based on the described four domains.
• Benchmarking domain generalization algorithms using the described clear daytime source and adverse target domain protocol.
• Studying the impact of weather and illumination on UAV-based computer vision tasks.

Strengths

• Provides annotations in the widely-used COCO format.
• Explicitly structures data into four distinct weather and illumination domains for controlled evaluation.
• Built from the established VisDrone/DroneVehicle dataset as a source.

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

Derived from the original VisDrone/DroneVehicle dataset.

Collection Method

Reorganized and annotated for domain-generalization benchmarking.

Freshness

Last updated 2026-05-22 14:38:04