论文标题
关键框架提案网络，用于视频中有效姿势估算
Key Frame Proposal Network for Efficient Pose Estimation in Videos
论文作者
论文摘要
人类姿势估计在视频中依赖于本地信息，这是通过独立估算每个框架或在跨帧的跟踪姿势来依赖于本地信息。在本文中，我们提出了一种新颖的方法，将本地方法与全球环境相结合。我们介绍了一个轻巧的，无监督的，关键的框架提案网络（K-FPN），以选择信息丰富的框架和学习的词典，以从这些帧中恢复整个姿势序列。 K-FPN加快了姿势估计，并通过遮挡，运动模糊和照明变化为不良框架提供了鲁棒性，而学习的字典则提供了全局动态环境。 Penn Action和Sub-JHMDB数据集的实验表明，该提出的方法可以实现最先进的准确性，并具有很大的加快。
Human pose estimation in video relies on local information by either estimating each frame independently or tracking poses across frames. In this paper, we propose a novel method combining local approaches with global context. We introduce a light weighted, unsupervised, key frame proposal network (K-FPN) to select informative frames and a learned dictionary to recover the entire pose sequence from these frames. The K-FPN speeds up the pose estimation and provides robustness to bad frames with occlusion, motion blur, and illumination changes, while the learned dictionary provides global dynamic context. Experiments on Penn Action and sub-JHMDB datasets show that the proposed method achieves state-of-the-art accuracy, with substantial speed-up.
