Key Frame Proposal Network for Efficient Pose Estimation in Videos
This method improves pose estimation efficiency and robustness for video analysis applications, though it is incremental as it builds on existing local and global approaches.
The paper tackles efficient human pose estimation in videos by proposing a key frame proposal network (K-FPN) to select informative frames and a learned dictionary to recover poses, achieving state-of-the-art accuracy with substantial speed-up on Penn Action and sub-JHMDB datasets.
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.