PhyCoVIS: A visual analytic tool of physical coordination for cheer and dance training

TLDR

Dance video analysis is essential for online cheer and dance training to assess physical coordination. We propose a visualization‑driven approach to analyze dance videos. The system encodes video frames into heat maps using a neural‑network pose estimator, computes pose similarity to quantify differences from a standard, and offers an interactive visualization for temporal analysis. Case studies demonstrate the tool’s effectiveness in supporting physical coordination research.

Abstract

Abstract Processing and analyzing dance videos are important in the application of online cheer leading and dance training for physical coordination measurement. However, it is challenging for users to evaluate a massive amount of uploaded video, to precisely quantize and compare dance moves, and to visualize training results. To overcome these challenges, we propose a visualization‐driven approach for analyzing dance videos. We first encode extracted video frames into a set of heat maps via neural network, which calculates a skeleton structure for pose estimation with enhanced post‐processing to help capture dance moves. A subsequent pose similarity method allows users to quantize differences between student training videos and the standard one. Finally, an interactive visualization tool enables users and domain experts to interactively analyze the quality of dance moves along the time line. We demonstrate the applicability and effectiveness of our proposed tool using case studies involving physical coordination research.