In human pose estimation, a comprehensive evaluation of state-of-the-art frameworks is necessary to advance both research and practical applications. This paper presents a thorough review of state-of-the-art 2D and 3D human pose estimation frameworks, analyzing 118 papers and four GitHub repositories, with a focus on frameworks made since 2019. The following frameworks are chosen based on predefined inclusion criteria: AlphaPose, Detectron2, MediaPipe, MeTRAbs, MHFormer, MMPose, MoveNet, OpenPifPaf, OpenPifPaf-vita, OpenPose, PoseFormerV2, rtmlib, StridedTransformer-Pose3D, ultralytics (YOLOv8), ViTPose, and YOLOv7. This paper evaluates these 16 frameworks on an existing, unpublished dataset consisting of exercise videos recorded with a monocular RGB camera and synchronized gold-standard motion capture data. The dataset includes videos of nine individuals performing eight exercises, recorded from two camera views with different planar angles. The analysis evaluates joint angle performance of the frameworks using weighted mean absolute error and weighted intraclass correlation coefficient as quantitative metrics. MeTRAbs emerged as the best overall framework, while AlphaPose, rtmlib, and YOLOv7 were the top 2D performers.

Visual impairment is a widespread global health issue that affects millions of people across all ages and backgrounds. Timely intervention is essential for the effective management of eye diseases. Smartphones offer the possibility of continuously recording facial gestures during interaction with the device, whereby changes such as squinting of the eyes could indicate progressive vision loss. In this context, a mobile health application was developed to conduct a digital logMAR test while simultaneously capturing real-time facial features. A total of 37 participants took part in a controlled mobile eye test study. The facial landmarks recorded during the test were analyzed to identify patterns that can distinguish between sequences of letters that were read correctly, partially, or not at all. Specifically, explorative data analysis and receiver operating characteristic curves were employed to determine facial landmarks with high discriminative power in relation to reading ability. The predominant facial regions that showed the most significant change under reduced performance during the vision test were the nose, mouth, and cheeks. Notably, the characteristic maximum squinting of the cheeks stood out with an area under the curve of 0.82. The analysis showed the potential of tracking specific facial features for continuous and unobtrusive vision assessment. It motivates to integrate facial feature analysis into an everyday application such as a web browser and to conduct a study in a non-standardized environment on a larger scale.