Book contents
- Monitoring Laws
- Monitoring Laws
- Copyright page
- Contents
- Acknowledgements
- 1 Monitoring Laws
- 2 The Image and Institutional Identity
- 3 Images and Biometrics – Privacy and Stigmatisation
- 4 Dossiers, Behavioural Data, and Secret Speculation
- 5 Data Subject Rights and the Importance of Access
- 6 Automation, Actuarial Identity, and Law Enforcement Informatics
- 7 Algorithmic Accountability and the Statistical Legal Subject
- 8 From Photographic Image to Computer Vision
- 9 Person, Place, and Contest in the World State
- 10 Law and Legal Automation in the World State
- Index
8 - From Photographic Image to Computer Vision
Neural Networks and Identity in the World State
Published online by Cambridge University Press: 08 November 2019
- Monitoring Laws
- Monitoring Laws
- Copyright page
- Contents
- Acknowledgements
- 1 Monitoring Laws
- 2 The Image and Institutional Identity
- 3 Images and Biometrics – Privacy and Stigmatisation
- 4 Dossiers, Behavioural Data, and Secret Speculation
- 5 Data Subject Rights and the Importance of Access
- 6 Automation, Actuarial Identity, and Law Enforcement Informatics
- 7 Algorithmic Accountability and the Statistical Legal Subject
- 8 From Photographic Image to Computer Vision
- 9 Person, Place, and Contest in the World State
- 10 Law and Legal Automation in the World State
- Index
Summary
In 2012, Alex Krizhevsky, then a PhD student at University of Toronto under Geoffrey Hinton, won the annual ‘ImageNet’ image labelling competition by an impressive 10.8 per cent margin. His use of a neural network-based object classification algorithm would then trigger a major shift the way computers would relate to images and the physical world more generally. ImageNet is an image database first published by computer scientist Fei-Fei Li in 2009 and labelled primarily by Amazon Mechanical Turk workers. Its intention was to ‘map out the entire world of objects’ for the sake of training machine learning systems. The first winner of the ImageNet competition in 2010 achieved a labelling accuracy of 71.8 per cent.
- Type
- Chapter
- Information
- Monitoring LawsProfiling and Identity in the World State, pp. 135 - 157Publisher: Cambridge University PressPrint publication year: 2019