SPARK 2021

The SPARK dataset is a unique space multi-modal annotated image dataset containing a total of ~150k RGB images and the same number of depth images of 11 object classes (10 spacecrafts and one class of space debris).

SPARK 2022

The SPARK 2022 dataset contains two streams of data. 1. Spacecraft Detection involves localising and classifying the object. 2. Spacecraft Trajectory Estimation focuses on temporal data to estimate the 6DoF pose of the spacecraft.

SPARK 2024

The SPARK 2024 dataset contains two streams of data. Stream 1 – Spacecraft Semantic Segmentation Segmenting space objects: classify pixel values based on which part of the object they belong to. Stream 2 – Spacecraft Trajectory Estimation Leveraging the knowledge of temporal data to estimate the 6DoF pose of the spacecraft.

SPARK 2026

The SPARK 2026 dataset contains two streams of data. Stream 1 – Spacecraft Semantic Segmentation Segmenting space objects: classify pixel values based on which part of the object they belong to. Stream 2 – Spacecraft Trajectory Estimation Leveraging the knowledge of temporal data to estimate the 6DoF pose of the spacecraft.

SPADES

SPADES (SPAcecraft Pose Estimation Dataset using Event Sensing) is an event-sensing dataset for spacecraft pose estimation tasks and it uses the Proba-2 satellite of the PROBA-2 mission as the target. SPADES dataset includes simulated and real event data obtained from a realistic satellite model at the SnT ZeroG testbed facility.

ECLIPSE

ECLIPSE, the first benchmark designed to isolate illumination as an independent experimental variable. ECLIPSE pairs a large-scale photorealistic synthetic training set with a fully labeled HIL evaluation set of 14 approach trajectories, each replayed identically under 4 controlled lighting conditions, enabling rigorous evaluation of illumination-robust and domain-adaptation methods. Through this controlled design, our experiments confirm that lighting is the dominant driver of the sim-to-real performance gap in spacecraft pose estimation.