Open-Source Biomechanical Video Analysis and Comparison Dashboard

Abstract:
This dataset provides an open-source, interactive web application framework built using Streamlit and MediaPipe for markerless biomechanical motion comparison. Designed for sports scientists, coaches, and researchers, the "Motion Comparison Dashboard" enables side-by-side video analysis of two separate athletes or multiple trials of the same individual.

The system automates pose extraction using MediaPipe's deep learning models to dynamically calculate 2D joint angles (knees, hips, elbows), vertical jump height, airtime, step rate, and horizontal ankle trajectories. To address the limitations of pixel-space metrics, the application includes a calibration module that approximates real-world metric units (meters and meters per second) based on the subject's baseline height. The repository includes source code, structural dependencies, and data visualization pipelines utilizing Plotly (time-series charts and normalized radar charts). This tool offers an accessible, low-cost alternative to expensive laboratory-grade motion capture systems for field-based athletic kinematic evaluations.

Key Words: Movement Analysis, Open CV, Streamit