Table of Contents
Active Perceptual Interface using hidden-state reinforcement
learning
Overview
Vision as Interface
Body tracking interfaces - ALIVE
The ALIVE System
ALIVE overview
Body-tracking routine
Foreground color models
Scene projection
Video: Body tracking and ALIVE
Active Face and Gesture Tracking
Gesture / Expression tracking
PPT Slide
PCA-based View class model
PPT Slide
PPT Slide
Facial Tracking: Hardware
Our approach
Model mimics user (in real time)
Video: Expression
Overview
Active Vision framework
Visual Attention for Active Interface
Attention for Recognition
Perceptual action-selection
POMDPs for Visual Attention
Active Recognition
Active Recognition Example
Active Recognition Example Domain
Learning Active Recognition
Q-learning
Hidden-state Q-Learning
Memory-based Q-function
Nearest Sequence Matching
Memory-based Q-learning
Variable K-NN
Multiple-target AGR Example
Active Recognition Example - 1
Active Recognition Example - 2
Active Recognition Example - 3
Recognition results
Recognition results - the fine print
Extracting concise behavior
Sample extracted recognition routine
State-chain representation
Integration of multiple chains
Merge redundant steps
Overview
Perceptive Network Interfaces
Attentive Avatars
Two rendering approaches
Future Work |