"Dataset Dataset for Space Target Recognition, Segmentation, and Pose Estimation Based on Simulated ISAR Images"

"This dataset provides a comprehensive set of synthetic Inverse Synthetic Aperture Radar (ISAR) images for spaceborne satellite target recognition, semantic segmentation, and pose estimation tasks. It contains 6,912 radar images, categorized into four satellite classes (AcrimSAT, AIM, Aquarius, and Starlink), with a resolution of 128\u00d7128 pixels per image. Each satellite is simulated using a point scatterer model and processed via ISAR Fast Fourier Transform (FFT) to generate range\u2013Doppler images corresponding to different azimuth and elevation angles.The dataset includes pixel-level segmentation masks for each image, as well as detailed pose annotations, including XYZ-axis rotation angles sampled at 30-degree intervals. This provides a valuable resource for multi-task learning algorithms and remote sensing applications. Although the dataset is entirely synthetic, its construction is designed to maintain relevance to real-world satellite recognition and pose estimation tasks.This dataset aims to promote further research in radar signal processing, machine learning for remote sensing, and multi-task learning for spaceborne satellite applications, and can be used for developing and benchmarking new models for recognition, segmentation, and pose estimation tasks."