Jump to content
Visual Motion Processing / Titelei/Inhaltsverzeichnis
Visual Motion Processing / Titelei/Inhaltsverzeichnis
Contents
Chapter
Expand
|
Collapse
Page
Titelei/Inhaltsverzeichnis
Details
1–11
1 Introduction
1–11
Details
1.1 Biological Motion Recognition
Details
1.1.1 Temporal and View-Point Variations
Details
1.1.2 Discriminative Features
Details
1.2 Computational Models for Biological Motion Recognition
Details
1.2.1 Computational Neuroscience
Details
1.3 Summary & Thesis Structure
Details
12–25
2 Computational Model
12–25
Details
2.1 Biological Motion Recognition in the Brain
Details
2.1.1 Different Motion Representations
Details
2.1.2 Static and Dynamic Form Description
Details
2.1.3 Neurophysiological Experiments
Details
2.1.4 Brain Areas based on Neurophysiological Experiments
Details
2.1.5 Dorsal Stream Areas In Biological Motion Recognition
Details
2.1.6 Mid-Level Motion Patterns
Details
2.1.7 Ventral Stream Areas In Biological Motion Recognition
Details
2.1.8 Posterior Superior Temporal Sulcus (STSp)
Details
2.2 Proposed Computational Model
Details
2.2.1 Feed-Forward Neural Networks
Details
2.2.2 Invariance Properties of Feed-Forward Neural Networks
Details
2.2.3 RelatedWork
Details
2.2.4 Proposed Computational Model
Details
26–57
3 Unsupervised Pattern Learning
26–57
Details
3.1 RelatedWork
Details
3.1.1 Principal Component Analysis
Details
3.1.2 Independent Component Analysis
Details
3.1.3 Extensions of NMF
Details
3.2 Properties of Parts-based Representations
Details
3.2.1 Basic Constraints
Details
3.2.2 Non-negativity
Details
3.2.3 Sparsity
Details
3.2.4 Local and Lateral Inhibition
Details
3.2.5 Resulting Energy Function and Notations
Details
3.3 Sparse Non-negative Matrix Factorization
Details
3.3.1 Sparse Activations
Details
3.3.2 Normalized Basis Vectors
Details
3.3.3 Sparse Basis Vectors
Details
3.3.4 Reconstruction Energy
Details
3.3.5 sNMF Learning Algorithm
Details
3.3.6 Orthogonality and Enforced Parts-Basedness
Details
3.4 Non-negative Representations of Real-valued Data
Details
3.4.1 Multidimensional Input
Details
3.4.2 Multidimensional Basis Vectors
Details
3.4.3 Multidimensional Activations
Details
3.4.4 Sparse Activation Amplitudes
Details
3.4.5 Positive and Negative Input
Details
3.4.6 Strict Non-negativity
Details
3.4.7 Weak Non-negativity
Details
3.4.8 Orthogonality between Positive and Negative Reconstructions
Details
3.5 Translation-invariant NMF
Details
3.5.1 Reconstruction Energy
Details
3.5.2 Sparse Activations
Details
3.5.3 Orthogonality between Positive and Negative Representation
Details
3.5.4 Enforced Topological Sparsity
Details
3.5.5 VNMF Learning Algorithm
Details
3.6 AlgorithmSummary
Details
58–88
4 Optical Flow Estimation
58–88
Details
4.1 ProblemFormulation
Details
4.1.1 General Algorithmic Approaches
Details
4.1.2 Correlation Methods
Details
4.1.3 Differential Methods
Details
4.1.4 Method Comparison
Details
4.2 RelatedWork
Details
4.2.1 Horn and Schunk
Details
4.2.2 Lukas and Kanade
Details
4.2.3 Extensions of the Classical Methods
Details
4.2.4 Multi-Scale Methods
Details
4.2.5 Other OFE-algorithms
Details
4.3 VNMF-OFE Approach
Details
4.3.1 Restrict Optical Flow Field to Model
Details
4.3.2 Enforced Non-Negativity
Details
4.3.3 Penalize Opposing Directions
Details
4.3.4 Sparse Activity Amplitudes
Details
4.3.5 Lateral Competition
Details
4.3.6 VNMF-OFE Learning Algorithm
Details
4.3.7 VNMF-OFE Algorithm for Activation Inference
Details
4.4 Learning the Basis Vectors
Details
4.4.1 Varying Model Parameters
Details
4.4.2 Varying Energy Parameters
Details
4.4.3 Learned vs Designed Basis Vectors
Details
4.4.4 Discussion of the Parameter Settings
Details
4.5 Comparison & Results
Details
4.5.1 Comparison to Related Work
Details
4.5.2 VNMF-OFE for Human Actions
Details
4.6 Summary & Discussion
Details
89–115
5 Feature Extraction
89–115
Details
5.1 Optical Flow Patterns
Details
5.1.1 Preprocessing
Details
5.1.2 Varying Energy Parameters
Details
5.1.3 Varying Basis Vector Parameters
Details
5.1.4 Detailed Analysis of the Learning Process
Details
5.1.5 Comparison to PCA and sNMF
Details
5.1.6 Basis Vectors learned on Face Data
Details
5.2 Gradient Patterns
Details
5.2.1 Preprocessing
Details
5.2.2 Varying Energy Parameters
Details
5.2.3 Varying Basis Vector Parameters
Details
5.2.4 Detailed Analysis of the Learning Process
Details
5.2.5 Comparison to PCA and sNMF
Details
5.2.6 Basis Vectors learned on Face Data
Details
5.3 VNMF as Feature Descriptor
Details
5.3.1 Simple Cell Response
Details
5.3.2 Complex Cell Response
Details
5.3.3 Relation to HOG/HOF Descriptor
Details
116–125
6 Human Action Recognition
116–125
Details
6.1 Support Vector Machine (SVM)
Details
6.2 Results for Different Basis Vector Sets
Details
6.2.1 Varying Basis Vector Parameters
Details
6.2.2 Varying Energy Parameters
Details
6.2.3 Comparison to PCA and sNMF Patterns
Details
6.2.4 Varying Simple Cell Response
Details
6.3 Facial Expression Recognition
Details
6.4 Comparison to Related Work
Details
6.4.1 HOG/HOF Results
Details
6.4.2 Benchmark Results
Details
126–131
7 Conclusion
126–131
Details
7.1 Summary & Discussion
Details
7.1.1 Optical Flow Estimation (VNMF-OFE)
Details
7.1.2 Feature Extraction (VNMF)
Details
7.1.3 Biological Motion Recognition Model (FFNN)
Details
7.2 Outlook
Details
132–132
A Bag of Words
132–132
Details
133–134
B Visual Cortex
133–134
Details
135–136
C Gradient Derivations
135–136
Details
C.1 Translation Invariant Learning
Details
C.2 Topological Sparsity
Details
137–142
D Sparse Non-Negative Linear Dynamic Systems
137–142
Details
D.1 Temporal Extension of sNMF
Details
D.2 RelatedWork
Details
D.3 Transition Energy
Details
D.3.1 Sparsity in the Transitions
Details
D.3.2 sNN-LDS Learning Algorithm
Details
D.4 Results
Details
143–158
Bibliography
143–158
Details
Durchsuchen Sie das Werk
Geben Sie ein Keyword in die Suchleiste ein
Partial access
Visual Motion Processing
Titelei/Inhaltsverzeichnis
Autoren
Thomas Guthier
DOI
doi.org/10.51202/9783186251084-I
ISBN print: 978-3-18-525108-5
ISBN online: 978-3-18-625108-4
Chapter Preview
Share
Download PDF
Download citation
RIS
BibTeX
Copy DOI link
doi.org/10.51202/9783186251084-I
Share by email
Video schließen
Share by email VDI Verlag eLibrary
Recipient*
Sender*
Message*
Your name
Send message
This site is protected by reCAPTCHA and the Google
Privacy Policy
and
Terms of Service
apply.