Deep Learning to See : Towards New Foundations of Computer Vision.
Betti, Alessandro.
Deep Learning to See : Towards New Foundations of Computer Vision. - 1st ed. - 1 online resource (116 pages) - SpringerBriefs in Computer Science Series . - SpringerBriefs in Computer Science Series .
Intro -- Preface -- Acknowledgements -- Contents -- 1 Motion Is the Protagonist of Vision -- 1.1 Introduction -- 1.2 The Big Picture -- 1.3 Supervised Learning Is an Artificial Learning Protocol -- 1.4 Cutting the Umbilical Cord with Pattern Recognition -- 1.5 Dealing with Video Instead of Images -- 1.6 Ten Questions for a Theory of Vision -- 2 Focus of Attention -- 2.1 Introduction -- 2.2 How Can Humans Perform Pixel Semantic Labeling? -- 2.3 Insights from Evolution of the Animal Visual System -- 2.4 Why Focus of Attention? -- 2.5 What Drives Eye Movements? -- 2.6 The Virtuous Loop of Focus of Attention -- 3 Principles of Motion Invariance -- 3.1 Introduction -- 3.2 Computational Models in Spatiotemporal Environments -- 3.3 Object Identity and Affordance -- 3.4 From Material Points to Pixels -- 3.5 The Principle of Material Point Invariance -- 3.6 The Principle of Coupled Motion Invariance -- 3.7 Coupling of Vision Fields -- 4 Foveated Neural Networks -- 4.1 Introduction -- 4.2 Why Receptive Fields and Hierarchical Architectures? -- 4.3 Why Two Different Mainstreams? -- 4.4 Foveated Nets and Variable Resolution -- 5 Information-Based Laws of Feature Learning -- 5.1 The Simplest Case of Feature Conjugation -- 5.2 Neural Network Representation of the Velocity Field -- 5.3 A Dynamic Model for Conjugate Features and Velocities -- 5.4 Online Learning -- 5.5 Online Learning: An Optimal Control Theory Prospective -- 5.6 Why is Baby Vision Blurred? -- 6 Non-visual Environmental Interactions -- 6.1 Object Recognition and Related Visual Skills -- 6.2 What Is the Interplay with Language? -- 6.3 The ``en Plein Air'' Perspective -- Appendix A Calculus of Variations -- A.1 Integral Functional and Euler Equations -- Appendix References -- -- Index.
9783030909871
Computer vision.
Electronic books.
Deep Learning to See : Towards New Foundations of Computer Vision. - 1st ed. - 1 online resource (116 pages) - SpringerBriefs in Computer Science Series . - SpringerBriefs in Computer Science Series .
Intro -- Preface -- Acknowledgements -- Contents -- 1 Motion Is the Protagonist of Vision -- 1.1 Introduction -- 1.2 The Big Picture -- 1.3 Supervised Learning Is an Artificial Learning Protocol -- 1.4 Cutting the Umbilical Cord with Pattern Recognition -- 1.5 Dealing with Video Instead of Images -- 1.6 Ten Questions for a Theory of Vision -- 2 Focus of Attention -- 2.1 Introduction -- 2.2 How Can Humans Perform Pixel Semantic Labeling? -- 2.3 Insights from Evolution of the Animal Visual System -- 2.4 Why Focus of Attention? -- 2.5 What Drives Eye Movements? -- 2.6 The Virtuous Loop of Focus of Attention -- 3 Principles of Motion Invariance -- 3.1 Introduction -- 3.2 Computational Models in Spatiotemporal Environments -- 3.3 Object Identity and Affordance -- 3.4 From Material Points to Pixels -- 3.5 The Principle of Material Point Invariance -- 3.6 The Principle of Coupled Motion Invariance -- 3.7 Coupling of Vision Fields -- 4 Foveated Neural Networks -- 4.1 Introduction -- 4.2 Why Receptive Fields and Hierarchical Architectures? -- 4.3 Why Two Different Mainstreams? -- 4.4 Foveated Nets and Variable Resolution -- 5 Information-Based Laws of Feature Learning -- 5.1 The Simplest Case of Feature Conjugation -- 5.2 Neural Network Representation of the Velocity Field -- 5.3 A Dynamic Model for Conjugate Features and Velocities -- 5.4 Online Learning -- 5.5 Online Learning: An Optimal Control Theory Prospective -- 5.6 Why is Baby Vision Blurred? -- 6 Non-visual Environmental Interactions -- 6.1 Object Recognition and Related Visual Skills -- 6.2 What Is the Interplay with Language? -- 6.3 The ``en Plein Air'' Perspective -- Appendix A Calculus of Variations -- A.1 Integral Functional and Euler Equations -- Appendix References -- -- Index.
9783030909871
Computer vision.
Electronic books.