Datasets

In this satellite/space object detection (SOD) dataset, we have modeled a dynamic solar system with accurate celestial scales and distances. LEO satellites were placed at altitudes between 500 km and 600 km with random orbital configurations. Each satellite was equipped with an onboard camera with a fixed 45-degree field of view (FoV), which sequentially captured images as the satellite moved through its orbit. The camera script automatically recorded distances and annotated bounding boxes for satellites detected within a 5 km range.

This satellite clustering for SOD (SCS) dataset is designed to simulate scenarios involving multiple nearby satellites capturing the same scene from slightly offset positions but identical viewing angles. The goal is to study how a multi-view setting influences satellite object detection. We use a fixed cluster size at three satellites in this dataset.