SyncUp: Vision-based Practice Support for Synchronized Dancing
summary
SyncUp is a vision system that analyzes practice videos with off‑the‑shelf pose estimation, computes Body-part-level Pose Distance (BPD) and summarizes it into framewise overall pose similarity (OPS), and builds an Impact Envelope signal for each dancer to run cross‑correlation and detect who is ahead/behind. It generates heatmap overlays, 1‑D timeline heatmaps and click‑to‑jump navigation, is tested on 18 practice videos from five groups with ~200 human ratings, and its pose and temporal predictions correlate strongly with those ratings.
AI-written summary · about
related talks
30mPoseDance: Build a TikTok Trainer
28mVibing Vision: how we built a pose tracking art installation in two weeks
36mDancing With Myself
29mOf Muppets, Metaballs, and Ballroom Dance!
10mPoseTween: Pose-driven Tween Animation
02mRecap Video
39mChoreorobotics: Teaching Robots How To Dance
01h 15mPre-sets and Post-sets: Poetics and Projections