JMIR Publications

ABSTRACT

Progress in robot manipulation has recently been driven less by algorithmic breakthroughs than by the sheer scale of real-world demonstration data that has become cheap to collect. Locomotion and everyday movement analysis have not had an equivalent moment: most gait research still depends on instrumented laboratories, marker-based motion capture, or proprietary clinical hardware, which puts objective movement data out of reach for most people and most researchers. This paper introduces the seed release (v0.1) of an open, privacy-preserving approach to markerless gait analysis built entirely from smartphone video. We describe the underlying capture-and-scoring pipeline, report descriptive statistics from an early, self-selected pilot cohort, and set out a governance roadmap for a properly consented, versioned, clip-level benchmark intended for release in future versions. We frame this early release deliberately as a beginning rather than a finished dataset, and invite researchers, clinicians, and the broader pose-estimation and robotics community to shape what a shared, open movement benchmark should look like.