arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1603.04037cs.CV
收藏

Pose for Action(动作)

Pose for Action - Action for Pose

University of Bonn(波恩大学)

Umar Iqbal, Martin Garbade, Juergen Gall

摘要:

在这项工作中,我们建议利用人类行动的信息,以提高单目视频的姿态估计。为此,我们提出了一个图形结构模型,利用高层次的活动信息,将高阶部分的依赖关系,通过建模动作特定的外观模型和构成先验。然而,我们的姿势估计框架有效地估计了动作先验,而不是使用额外的昂贵的动作识别框架。这是通过从统一的动作先验开始并在姿态估计期间更新动作先验来实现的。我们还表明,学习适量的动作类之间的外观共享提高了姿态估计。我们在两个具有挑战性的数据集上证明了所提出的方法的有效性,这些数据集用于姿势估计和动作识别,具有超过80,000张测试图像。

英文摘要:

In this work we propose to utilize information about human actions to improve pose estimation in monocular videos. To this end, we present a pictorial structure model that exploits high-level information about activities to incorporate higher-order part dependencies by modeling action specific appearance models and pose priors. However, instead of using an additional expensive action recognition framework, the action priors are efficiently estimated by our pose estimation framework. This is achieved by starting with a uniform action prior and updating the action prior during pose estimation. We also show that learning the right amount of appearance sharing among action classes improves the pose estimation. We demonstrate the effectiveness of the proposed method on two challenging datasets for pose estimation and action recognition with over 80,000 test images.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑