Tilting at windmills: Data augmentation for deep pose estimation does not help with occlusions

TL;DR

This paper demonstrates that current deep pose estimation methods are highly vulnerable to occlusions, and that data augmentation techniques do not effectively improve robustness against occlusion challenges.

Contribution

The study introduces targeted occlusion attacks and evaluates their impact, revealing the limited effectiveness of data augmentation in addressing occlusion issues in pose estimation.

Findings

Pose estimation methods are not robust to occlusion.

Data augmentation does not significantly improve occlusion robustness.

Targeted occlusion attacks reveal vulnerabilities in current models.

Abstract

Occlusion degrades the performance of human pose estimation. In this paper, we introduce targeted keypoint and body part occlusion attacks. The effects of the attacks are systematically analyzed on the best performing methods. In addition, we propose occlusion specific data augmentation techniques against keypoint and part attacks. Our extensive experiments show that human pose estimation methods are not robust to occlusion and data augmentation does not solve the occlusion problems.