The making of a neuromorphic visual system / by Christoph Rasche.

Includes bibliographical references (p. [129]-135) and index.

1 Seeing: Blazing Processing Characteristics 1
1.1 An Infinite Reservoir of Information 1
1.2 Speed 1
1.3 Illusions 2
1.4 Recognition EJvoIvement 2
1.5 Basic-Level Categorization 3
1.6 Memory Capacity and Access 5
1.7 Summary 5
2 Category Representation and Recognition Rvolvement 7
2.1 Structural Variability Independence 7
2.2 Viewpoint Independence 7
2.3 Representation and Evolvement 9
2.3.1 Identification Systems 10
2.3.2 Part-based Descriptions 11
2.3.3 Template Matching 13
2.3.4 Scene Recognition 13
2.4 Recapitulation 14
2.5 Refining the Prlmaiy Engineering Goal 15
3 Neuroscientific Inspiration 17
3.1 Hierarchy and Models 17
3.2 Criticism and Variants 20
3.3 Speed 23
3.4 Alternative 'Codes' 25
3.5 Alternative Shape Recognition 27
3.6 Insight from Cases of Visual Agnosia 29
3.7 Neuronal Level 31
3.8 Recapitulation and Conclusion 35
4 Neuromorphic Tools 37
4.1 The Transistor 37
4.2 A Synaptlc Circuit 38
4.3 Dendritic Compartments 39
4.4 An Integrate-and-Flre Neuron 40
4.5 A Silicon Cortex 40
4.6 Fabrication Vagrancies require Simplest Models 42
4.7 Recapitulation 42
5 Insight From Line Drawings Studies 45
5.1 A Representation with Polygons 45
5.2 A Representation wdth Polygons and their Context .. . . 49
5.3 Recapitulation 51
6 Retina Circuits Signaling and Propagating Contours 55
6.1 The Input; a Luminance Landscape 55
6.2 Spatial Analysis in the Real Retina 55
6.2.1 Method of Adjustable Thresholds 57
6.2.2 Method of Latencies 58
6.3 The Propagation Map 58
6.4 Signaling Contours in Gray-Scale Images 60
6.4.1 Method of Adjustable Thresholds 60
6.4.2 Method of Latencies 60
6.4.3 Discussion 64
6.5 Recapitulation 64
7 The Symmetric-Axis Transform 67
7.1 The Transform 67
7.2 Architecture 68
7.3 Performance 70
7.4 SAT Variants 74
7.5 Fast Waves 74
7.6 Recapitulation 75
8 Motion Detection 77
8.1 Models 77
8.1.1 Computational 77
8.1.2 Biophysical 77
8.2 Speed Detecting Architectures 79
8.3 Simulation 81
8.4 Biophysical Plausibility 83
8.5 Recapitulation 85
9 Neuromorphic Architectures: Pieces and Proposals 87
9.1 Integration Perspectives 87
9.2 Position and Size Invariance 89
9.3 Architecture for a Template Approach 92
9.4 Basic-Level Representations 94
9.5 Recapitulation 95
10 Shape Recognition with Contour Propagation Fields 97
10. IThe Idea of the Contour Propagation Field 97
10.2 Architecture 98
10.3Testing jOO
10.4 Discussion 104
10.5Learning. . . 1Q7
10.6 Recapitulation 109
11 Scene Recognition 111
11.1 Objects in Scenes, Scene Regularity Ill
11.2Representation. Evolvement, Gist Ill
11.3 Scene Exploration 113
11.4 Engineering 115
11.5 Recapitulation 116
12 Summary 117
12.1 The Quest for Efficient Representation and Elvolvement . 117
12.2Contour Extraction and Grouping 121
12.3Neuroscientific Inspiration 121
12.4Neuromorphic Implementation 122
12.5 Future Approach 122