Improved 2D Keypoint Detection in Out-of-Balance and Fall Situations -- combining input rotations and a kinematic model
This work addresses injury analysis in alpine skiing, providing a domain-specific dataset and method, but it is incremental as it builds on existing pose estimation techniques.
The paper tackled the problem of 2D keypoint detection in challenging scenarios like falls and out-of-balance situations by proposing a post-processing routine that combines input rotations with a kinematic model, resulting in an improvement of up to 21% in detection accuracy as measured by PCK@0.2.
Injury analysis may be one of the most beneficial applications of deep learning based human pose estimation. To facilitate further research on this topic, we provide an injury specific 2D dataset for alpine skiing, covering in total 533 images. We further propose a post processing routine, that combines rotational information with a simple kinematic model. We could improve detection results in fall situations by up to 21% regarding the PCK@0.2 metric.