270 likes | 413 Views
Snap Together Motion : Assembling Run-Time Animation. Michael Gleicher Hyun Joon Shin Lucas Kovar Andrew Jepsen University of Wisconsin. Virtual Environment. demand on motion synthesis motion quality control little run time overload. Virtual Environment. demand on motion synthesis
E N D
Snap Together Motion : Assembling Run-Time Animation Michael Gleicher Hyun Joon Shin Lucas Kovar Andrew Jepsen University of Wisconsin
Virtual Environment • demand on motion synthesis • motion quality • control • little run time overload
Virtual Environment • demand on motion synthesis • motion quality • control • little run time overload • snap together motion
Basic Idea • processing a corpus of motion data • constructing a highly connected motion graph • making motion segments “snappable”
Basic Idea • processing a corpus of motion data • quality • constructing a highly connected motion graph • control • making motion segments “snappable” • little runtime overhead
Motion Graph • Kovar et al., Lee et al., and Arikan et al. • constructing graph with edges of pieces of motion • graph searching to synthesize motion • smooth transitions in runtime
Overview • structured motion graph with hub nodes • structured graph • hub node: a node with a large number of edges • snappable motion segments • displacement map Original Motion Snappable Motion Hub Node
Overview • at run-time • a highly connected graph • concatenating motion pieces at edges … …
Authoring • graph construction • choosing match frames • snappable motions • average posture • displacement map • handling constraints
Choosing Match Frames • graph construction by user control/heuristic • finding a pose having largest match set • match set: poses with small errors • error metric: point cloud matching [Kovar et al] match set hub error matrix
Averaging Poses • coordinate frame alignment • align every match set into a coordinate frame • 2D rigid transformation that minimizes point cloud error point clouds two sequences of poses transformed
Averaging Poses • averaging joint angle • use quaternion estimator [Park, Shin, Shin] • orientation with the minimum squared sum of error
Displacement Maps • displacement map • fitting average velocities together with pose • two-level displacement map • coarse knots for fitting the pose • denser knots for fitting the velocity
Constraints • enforcing consistency in constrained position • problem: • solution: • collecting hub nodes with possible constraint conflicts • determining the constraint position over the set another hub node a hub node a piece with inconsistent constraint
Runtime Motion Synthesis • a motion piece • corresponding nodes + coordinate transform • graph traversal • user-controlled traversal • or search algorithm • motion synthesis • concatenating pieces • applying coordinate transformation • little overload
Results • Video
Summary • a motion synthesis frame work • controllability with highly connected graph • preprocessing everything in authoring step • assembling snappable clips • authoring tool • user-controllable graph construction • seamlessly snappable clips • preserving original motion quality
Contents • Introduction • Motion Graph • Overview • Choosing Match Frames • Transition • Matching Coordinate Frames • Displacement maps • Constraints • at Run Time