Wild6D
收藏资源简介:
Wild6D是由加州大学圣地亚哥分校的研究人员收集的一个大规模RGBD视频数据集,专注于类别级6D对象姿态估计。该数据集包含超过1.1百万张图像,覆盖1722个不同的对象实例和5个类别,如瓶子、碗、相机、笔记本电脑和杯子。Wild6D的创建旨在解决在复杂场景中对多样对象进行6D姿态估计的挑战,通过使用半监督学习方法和无需任何真实数据上的3D标注的新模型RePoNet,实现了在野外的类别级6D对象姿态估计。
Wild6D is a large-scale RGBD video dataset collected by researchers from the University of California, San Diego, focusing on category-level 6D object pose estimation. This dataset contains over 1.1 million images, covering 1722 distinct object instances across 5 categories including bottles, bowls, cameras, laptops, and cups. Wild6D was developed to address the challenge of 6D pose estimation for diverse objects in complex real-world scenes, and enables category-level 6D object pose estimation in the wild by leveraging semi-supervised learning approaches and the novel RePoNet model, which requires no 3D annotations on real-world data.




