PoseTween: Pose-driven Tween Animation
summary
PoseTween generates pose-driven tween animations by anchoring virtual objects to body joints or limbs and interpolating object offsets inside a local frame, using OpenPose for pose extraction and automatic association rules to pick driving joints or limb pairs. Missing or bad joints are filled and edits propagate with auto-regressive models and constrained smoothing solved by quadratic programming. A user study with 12 participants finds significantly fewer keyframes and less creation time, and implementation uses OpenPose as the pose estimator.
AI-written summary · about
chapters (8)
- 0:00Intro
- 0:13Motivation
- 1:38PoseTween
- 4:28User Interface
- 6:56Methodology - Inference of Associations • Preview driving joint(s)/limb
- 8:15Experiments - User Study • Exploitation tasks
- 9:115 Experiments - Pose Editing Tool Quantitative evaluation of the accuracy of our pose editing tool
- 9:51Applications
related talks
40mPoseDance Build a TikTok Trainer
30mPoseDance: Build a TikTok Trainer
07m#ProjectPuppetron
20mSecondary Motion for Performed 2D Animation
06mSyncUp: Vision-based Practice Support for Synchronized Dancing
28mBuilding Motion Controlled Websites in JavaScript
06mCoolMoves: User Motion Accentuation in Virtual Reality
20mMoSculp: Interactive Visualization of Shape and Time