Tilting at windmills: Data augmentation for deep pose estimation does not help with occlusions
This addresses a critical limitation in human pose estimation for computer vision applications, but the findings are incremental as they confirm existing concerns without providing a solution.
The paper tackled the problem of occlusion degrading human pose estimation performance by introducing targeted occlusion attacks and occlusion-specific data augmentation techniques, finding that pose estimation methods remain non-robust to occlusion and data augmentation fails to solve this issue.
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.