Benchmark RadarAI BENCHMARK PROFILE
SkySeaLand
SkySeaLand is a satellite object detection dataset with 1,307 high-resolution images and 19,101 bounding boxes across four classes. It provides COCO and YOLO annotations, a common split, and COCO metrics for detector evaluation.
- Released
- 2026-08-07
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
The benchmark addresses a gap in wide-format satellite imagery detection, offering a compact dataset with a fixed protocol to compare detectors under standard metrics, supporting practical model selection for transportation monitoring.
Motivation
Satellite object detection is challenged by small targets and wide-format scenes that lose detail under standard square-input resizing.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.