OTPose: Occlusion-Aware Transformer for Pose Estimation in Sparsely-Labeled Videos

TL;DR

OTPose introduces an occlusion-aware transformer framework that effectively estimates multi-human poses in sparsely-labeled videos, handling occlusion and motion blur with semi-supervised learning and temporal encoding.

Contribution

The paper presents a novel attention mask for occlusion handling and employs transformers for temporal dependency encoding in pose estimation from sparse video annotations.

Findings

State-of-the-art results on PoseTrack2017 and PoseTrack2018 datasets.

Robustness to occlusion and motion blur in sparse annotations.

Effective semi-supervised occlusion-aware heatmaps.

Abstract

Although many approaches for multi-human pose estimation in videos have shown profound results, they require densely annotated data which entails excessive man labor. Furthermore, there exists occlusion and motion blur that inevitably lead to poor estimation performance. To address these problems, we propose a method that leverages an attention mask for occluded joints and encodes temporal dependency between frames using transformers. First, our framework composes different combinations of sparsely annotated frames that denote the track of the overall joint movement. We propose an occlusion attention mask from these combinations that enable encoding occlusion-aware heatmaps as a semi-supervised task. Second, the proposed temporal encoder employs transformer architecture to effectively aggregate the temporal relationship and keypoint-wise attention from each time step and accurately refines the target frame's final pose estimation. We achieve state-of-the-art pose estimation results for PoseTrack2017 and PoseTrack2018 datasets and demonstrate the robustness of our approach to occlusion and motion blur in sparsely annotated video data.