← ArchivePaper2012

Tiling Motion Patches

Manmyung Kim, Youngseok Hwang, Kyunglyul Hyun, Jehee Lee

Motion synthesis approach using tiled motion patches for seamlessly transitioning and looping character locomotion in interactive applications.

Open the paper ↗Trace in the lineage graph

Abstract

Proposes a tiling algorithm for creating dense crowds of interacting virtual characters using deformable motion patches. The method collects episodes of multiple characters and tiles them spatially and temporally to generate seamless multi-character animation with complex interactions like hand shaking and object carrying, using a combination of stochastic sampling and deterministic search.

How to read this

Category: Method: multi-character motion synthesis
Contributions: Tiles deformable motion patches spatially and temporally to synthesize dense crowds of interacting characters
Collects episodes of multiple characters and assembles them into seamless multi-character animation
Combines stochastic sampling with deterministic search to generate complex interactions such as hand shaking and object carrying
Context: Builds on the motion-graph lineage (Kovar et al. Motion Graphs), extending data-driven reassembly from single-character paths to tiled multi-character interaction patches.Builds on: Motion Graphs
Correctness: Demonstrated on crowds with close interactions; the result space is bounded by the captured episodes and the tileability of patches, so interactions outside the recorded repertoire will not appear and seam quality depends on patch compatibility.
Clarity: Accessible; a first pass conveys the tiling metaphor, a second pass clarifies the sampling-plus-search synthesis loop.
How to read it: Focus on what a motion patch is and how tiles compose without seams; a second pass on the sampling/search algorithm is worth it if you are building interactive crowds.

Builds on

Motion Graphs 2002

Built upon by

Nothing yet.

Related work

Near-Optimal Character Animation with Continuous Control 2007 / SIGGRAPH
DeepPhase: Periodic Autoencoders for Learning Motion Phase Manifolds 2022 / SIGGRAPH
Physics-Based Character Controllers Using Conditional VAEs 2022 / SIGGRAPH
Automated Extraction and Parameterization of Motions in Large Data Sets 2004 / SIGGRAPH

Keywords

motion patches tiling algorithm multi-character interaction deformable patches motion capture mcmc reversible jump mcmc simulated annealing crowd simulation spatiotemporal constraints stitching patches motion deformation laplacian equations hand shaking carrying objects

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →