Computer Vision for Curious Machines
Symposium:
Session Title:
- Surveillant Spaces: From Autonomous Surveillance to Machine Voyeurism
Presentation Title:
- Computer Vision for Curious Machines
Presenter(s):
Venue(s):
Abstract:
Panel: Surveillant Spaces: From Autonomous Surveillance to Machine Voyeurism
Computer vision is a branch of artificial intelligence that is concerned with developing algorithms to allow machines to process and respond to visual data. The degree to which the early pioneers in artificial intelligence underestimated the challenges involved is often illustrated with an anecdote about Marvin Minsky instructing a student to solve “the problem of computer vision” as a summer project. The first breakthroughs in computer vision came from the field of computational neuroscience and the work of David Marr, building low-level models of the visual cortex from the ground up.
Nearly forty years later, the state-of-the-art in computer vision is still very much in the process of constructing relatively primitive representations from captured images. Nevertheless, the research has produced a wealth of techniques that can be applied to suitably structured scenes to extract meaningful information. Many of these techniques are simple enough that they can be implemented by novice programmers while more sophisticated techniques have become readily available through programming libraries and off-the-shelf software. The availability of computer vision technology provides a base for experimenting with machine autonomy in creative domains. In this panel I will discuss the possibility of developing autonomous machine performers that take advantage of the advances in computer vision by first reviewing some of relevant low-level and high-level techniques and showing how these can be integrated with machine learning systems. In particular, I will present this exploration in relation to my research developing self-motivated (curious) agents and my collaborations with Petra Gemeinboeck exploring the performativity of the gaze through the creation of machine augmented environments. In this workshop-like session we will explore the construction of a self-motivated machinic voyeur, examine what it sees, how it responds and what drives it.
