Summary & Homework Jinxiang Chai
Outline Motion data process paper summary Presentation tips Homework Paper assignment
Skeleton Model Extraction Estimate the skeleton model from motion capture data
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints)
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow)
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow) Motion edit (can satisfy user constraints and fast)
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow) Motion edit (can satisfy user constraints and fast) However: does not work for many user-defined constraints does not utilize spatial-temporal correlation in human motion
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow) Motion edit (can satisfy user constraints and fast) Expression cloning (facial data) (cannot satisfy user constraints)
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow) Motion edit (can satisfy user constraints and fast) Expression cloning (facial data) (cannot satisfy user constraints) Motion synopsis
Motion Data Processing I Operations on single motion sequence Motion warping (cannot satisfy user constraints) Motion retargeting (can satisfy user constraints but slow) Motion edit (can satisfy user constraints and fast) Expression cloning (facial data) (cannot satisfy user constraints) Motion synopsis What else?
Motion Data Processing I Operations on single motion sequence Motion segmentation? Motion recognition? Motion filtering? More operations in facial data
Motion Data Processing II Operations between two motion sequence Motion style translation What else?
Motion Data Processing II Operations between two motion sequence Motion style translation What else?
Motion Data Processing II Operations between two motion sequence Motion style translation Motion interpolation?
Motion Data Processing III Operations on multiple motion sequences/database Motion data compression
Motion Data Processing III Operations on multiple motion sequences/database Motion data compression What else?
Motion Data Processing III Operations on multiple motion sequences/database Motion data compression Motion Retrieval? Motion synthesis?
Presentation Tips Choose right background Use blue or black or dark colors Focus on important things Always talk about what, why, and how Detailize the most important technical content Always talk about the limitations of the paper How to deal with equations Avoid using too many equations Talk about the intuition of the equations Explain each term of the equations
Presentation Tips (Cont.) Choose the right background Use dark colors (blue or black) Be consistent Font type, size, color, bullets, figure captions etc. Use short & concise sentences Do not copy & paste from the paper Use images, diagrams, figures, videos to demonstrate ideas
Presentation Tips (Cont.) Play videos Explain each video Move the mouse away from the window
Example Slides
Goal: everyone can generate and control human motion easily and quickly Online animation control Interfaces for Controlling Human Motion
Applications: Online Animation Performance-based facial animation for home use Tiger Woods PGA Tour 2005 Mike Tyson Heavyweight Boxing Friday Night 3D Bowling Teleconferencing (from BT) Xbox Outlaw Tennis Multi-user Virtual Worlds Virtual Poker Room Virtual Presenter
Reordering motion clip [Lee et al. SIG02, Kovar et al. SIG02, Pullen & Bregler SIG02, Arikan et al. SIG03] Learning model from human motion [Brand & Hertzmann SIG00, Li et al. SIG02] Interpolating motions [Rose et al. CG&A98, Kovar et al. SIG04] Animation from Mocap Data
Video Analysis Overview Online motion synthesis Online local modeling Low-dimensional control signals Preprocessed motion capture data Preprocessed motion capture data Online local models
Online Motion Optimization Optimizing a nonlinear function in about 7 dimensional space w t : Control term Smoothness term Pose prior term … … ?
Outline Motion data paper summary Presentation tips Homework Paper assignment
Dimensionality reduction using PCA Current pose Mean pose Eigen-posesLow-dimensional space Linear model: Mean pose Eigen-poses Dim(w t )? Linear Dimensionality Reduction
A v e r a g e r e c o n s t r u c t i o n e r r o r ( d e g r e e p e r j o i n t a n g l e ) A large heterogeneous database (1 hour of data) PCA (38D for error<1 o ) Number of dimensions
Reconstruction Error Curve A v e r a g e r e c o n s t r u c t i o n e r r o r ( d e g r e e p e r j o i n t a n g l e ) A large database Number of dimensions of W t
Minimal Dimensionality A v e r a g e r e c o n s t r u c t i o n e r r o r ( d e g r e e p e r j o i n t a n g l e ) A large database Number of dimensions of W t dim7 dim
Videos Two Animation videos rendered by Maya Original mocap sequence Reconstructed mocap sequence
Startup Codes & Data Motion capture data (.amc files and.asf file) A skinned character model Visualize amc/asf file Read asf/amc file Visualize mocap data Forward kinematics Matlab codes Read and write.amc file
Softwares Maya To be installed in Rm 220, HRBB Render each frame based on mocap data and character model Instruction on how to render animation Adobe Premier Installed in Rm 220, HRBB Making video from image sequences
Outline Motion data paper summary Presentation tips Homework Paper assignment