SPEED+
收藏资源简介:
SPEED+是斯坦福大学航空航天系开发的下一代航天器姿态估计数据集,专注于跨领域差距。该数据集包含60,000张合成图像用于训练,以及9,531张硬件在环图像,这些图像是从TRON设施捕获的航天器模型。TRON是一个独特的机器人测试平台,能够捕获任意数量的目标图像,并提供准确且多样化的姿态标签以及高保真的航天器光照条件。SPEED+用于第二届国际卫星姿态估计挑战赛,旨在评估和比较基于合成图像训练的航天器机器学习模型的鲁棒性。数据集的应用领域包括未来的轨道服务和空间物流任务,旨在解决非合作目标的姿态确定和跟踪问题。
SPEED+ is a next-generation spacecraft attitude estimation dataset developed by the Department of Aeronautics and Astronautics of Stanford University, focusing on cross-domain gaps. This dataset contains 60,000 synthetic images for training, and 9,531 hardware-in-the-loop images of spacecraft models captured from the TRON facility. TRON is a unique robotic testbed that can capture an arbitrary number of target images, while providing accurate and diverse attitude labels as well as high-fidelity spacecraft lighting conditions. SPEED+ was used in the 2nd International Satellite Attitude Estimation Challenge, aiming to evaluate and compare the robustness of spacecraft machine learning models trained on synthetic images. The application domains of this dataset cover future on-orbit servicing and space logistics missions, aiming to address the issues of attitude determination and tracking for non-cooperative targets.

- 1SPEED+: Next-Generation Dataset for Spacecraft Pose Estimation across Domain Gap斯坦福大学航空航天系 · 2021年



