Health & Gait: a dataset for gait-based analysis.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Human Motion Analysis

Background:

• Gait metrics and anthropometric data are vital for assessing physical status.
• Current automated gait analysis methods use expensive, specialized equipment.
• There is a need for accessible, mobile-device-operable solutions.

Purpose of the Study:

• To introduce Health&Gait, the first dataset for video-based gait analysis.
• To provide a comprehensive dataset for developing and evaluating artificial vision algorithms for gait assessment.
• To facilitate automated patient monitoring and reduce healthcare professional burden.

Main Methods:

• Collected 1,564 videos from 398 participants.
• The dataset includes participant silhouettes, semantic segmentation, optical flow, and human pose.
• Data also contains anthropometric measurements (sex, height, weight) and gait parameters (step/stride length, gait speed).

Main Results:

• Technical evaluation demonstrated the utility of video-extracted information and gait parameters.
• Successfully performed tasks like sex classification and regression of weight and age.
• The dataset's richness supports diverse applications in automated gait analysis.

Conclusions:

• Health&Gait is a valuable resource for advancing video-based gait analysis.
• The dataset supports the development of accessible, mobile-first artificial vision systems for healthcare.
• Facilitates progress in automated physical status assessment and patient monitoring.