EurographicsSIGGRAPH Symposium on Computer Animation 2003 Unsupervised Learning
- Slides: 26
Eurographics/SIGGRAPH Symposium on Computer Animation (2003) Unsupervised Learning for Speech Motion Editing Yong Cao 1, 2 Petros Faloutsos 1 Frederic Pighin 2 University of California, Los Angeles 1 Institute for Creative Technologies, University of Southern California 2
Problem ■ Motion Capture is convenient but lacks flexibility ■ Problem: How to extract the semantics of the data for intuitive motion editing?
Related Work 1. Face motion synthesis ■ Physics-based face model Lee, Terzopoulos, Water Kähler, Haber, Seidel ( SIGGRAPH 1995) (Graphics Interface 2001) ■ Speech motion synthesis Bregler, Covell, Slaney Brand (SIGGRAPH 1997) (SIGGRAPH 1999) Ezzat, Pentland, Poggio (SIGGRAPH 2002) 2. Separation of style and content Brand, Hertzmann (SIGGRAPH 2000) Chuang, Deshpande, Bregler (Pacific Graphics 2002)
Our Contribution ■ New statistical representation of facial motion • Decomposition into style and content • Intuitive editing operations
Our Contribution Original Neutral Motion Edited Sad Motion
Roadmap ■ Independent Component Analysis (ICA) ■ Facial motion decomposition ■ Semantics of components ■ Motion editing
Independent Component Analysis (ICA) ■ Statistical technique ■ Linear transformation ■ Components are maximally independent
Steps of ICA ■ Preprocessing (PCA) ■ Centering ■ Whitening ■ ICA decomposition Reconstruction: Decomposition:
Roadmap ■ Independent Component Analysis (ICA) ■ Facial motion decomposition ■ Semantics of components ■ Motion editing
Speech motion Dataset Speech motion of 113 sentences in 5 emotion moods: Frustrated 18 sentences Happy 18 sentences Neutral 17 sentences Sad 30 sentences Angry 30 sentences Each motion: 109 motion capture markers 2 – 4 seconds
Facial Motion Decomposition and Reconstruction Facial motion Components in ICA space …………
Roadmap ■ Independent Component Analysis (ICA) ■ Facial motion decomposition ■ Semantics of components ■ Motion editing
Interpretation of independent components ■ Goal: Find the semantics of each component Classify each component into: Style (emotion) Content (speech) ■ Methodology ■ Qualitatively ■ Quantitatively
Qualitatively changing Style (emotion) Content (speech)
Quantitatively n Style: Emotion Same speech, different emotion ………… Happy Frustrated
Speech Content Grouping of motion markers ■ Mouth motion ■ Eyebrow motion ■ Eyelid motion
Content: speech related motion Step 1: Using each independent component to reconstruct facial motion 0 Reconstruct 0 0 ………… 0
Content: speech related motion Step 2: Compare according to certain region
Roadmap ■ Independent Component Analysis (ICA) ■ Facial motion decomposition ■ Semantic meaning of components ■ Motion editing
Motion Editing with ICA ■ Edit the motion in intuitive ways ■ Translate ■ Copy and Replace ■ Copy and Add
Results ■ Changing emotional state by translating
Conclusion ■ New statistical representation of facial motion • Decomposition into content and style • Intuitive editing operations
The End Thanks to Wen Tien for his help on this paper, Christos Faloutsos for useful discussions, and Brian Carpenter for his excellent performance. Thanks to the USC School of Cinema – Television and House of Moves for motion capture.
- "deep reinforcement learning"
- Data mining unsupervised learning
- Traditional animation vs computer animation
- Unsupervised learning in data mining
- Transductive learning for unsupervised text style transfer
- Autoencoders
- Maxnet neural network
- Workspca
- Unsupervised learning
- Andrew ng introduction to machine learning
- Www.youtube.com
- Autoencoder unsupervised learning
- Supervised and unsupervised learning
- Cuadro comparativo de e-learning
- Learning curves 2003
- Csc418
- Computer animation definition
- History of computer graphics
- General computer animation functions
- What is path based animation
- Unsupervised domain adaptation by backpropagation
- Unsupervised segmentation
- Object based image analysis
- Supervised vs unsupervised data mining
- Unsupervised models for named entity classification
- Unsupervised sentiment analysis using bert
- Unsupervised pos tagging