research papers

2025

1. 

   Health & Gait: a dataset for gait-based analysis

   NIH, Jan 2025

   Abs DOI Bib

   Acquiring gait metrics and anthropometric data is crucial for evaluating an individual’s physical status. Automating this assessment process alleviates the burden on healthcare professionals and accelerates patient monitoring. Current automation techniques depend on specific, expensive systems such as OptoGait or MuscleLAB, which necessitate training and physical space. A more accessible alternative could be artificial vision systems that are operable via mobile devices. This article introduces Health&Gait, the first dataset for video-based gait analysis, comprising 398 participants and 1, 564 videos. The dataset provides information such as the participant’s silhouette, semantic segmentation, optical flow, and human pose. Furthermore, each participant’s data includes their sex, anthropometric measurements like height and weight, and gait parameters such as step or stride length and gait speed. The technical evaluation demonstrates the utility of the information extracted from the videos and the gait parameters in tackling tasks like sex classification and regression of weight and age. Health&Gait facilitates the progression of artificial vision algorithms for automated gait analysis.

   @article{zafrapalmg2025healthgait,
     title = {Health \& Gait: a dataset for gait-based analysis},
     author = {},
     journal = {NIH},
     volume = {12},
     number = {44},
     pages = {44},
     year = {2025},
     month = jan,
     doi = {10.1038/s41597-024-04327-4},
     publisher = {Nature},
     url = {https://doi.org/10.1038/s41597-024-04327-4},
     keywords = {gait analysis, dataset, video-based analysis, pose estimation, anthropometry},
   }

2020

1. 

   HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation

   In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Mar 2020

   Abs DOI Bib PDF

   Bottom-up human pose estimation methods have diffi- culties in predicting the correct pose for small persons due to challenges in scale variation. In this paper, we present HigherHRNet: a novel bottom-up human pose estimation method for learning scale-aware representa- tions using high-resolution feature pyramids. Equipped with multi-resolution supervision for training and multi- resolution aggregation for inference, the proposed ap- proach is able to solve the scale variation challenge in bottom-up multi-person pose estimation and local- ize keypoints more precisely, especially for small person. The feature pyramid in HigherHRNet consists of feature map outputs from HRNet and upsampled higher-resolution outputs through a transposed convolution. HigherHR- Net outperforms the previous best bottom-up method by 2.5% AP for medium person on COCO test-dev, show- ing its effectiveness in handling scale variation. Further- more, HigherHRNet achieves new state-of-the-art result on COCO test-dev (70.5% AP) without using refinement or other post-processing techniques, surpassing all existing bottom-up methods. HigherHRNet even surpasses all top- down methods on CrowdPose test (67.6% AP), suggest- ing its robustness in crowded scene. The code and mod- els are available at https://github.com/HRNet/ Higher-HRNet-Human-Pose-Estimation.

   @inproceedings{cheng2020higherhrnet,
     title = {HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation},
     booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
     author = {},
     pages = {1--10},
     numpages = {10},
     year = {2020},
     month = mar,
     day = {12},
     doi = {10.1109/CVPR42600.2020.00543},
     url = {https://arxiv.org/abs/1908.10357},
     dimensions = {true},
   }