The Selective Tuning Attentive Reference model (STAR) is previewed (Tsotsos 2011, Tsotsos & Kruijne 2014). STAR, currently in its early stages of development, integrates proposals for an executive controller, working memory, attentional mechanisms, eye movements, visual processing and the communication among these that is hoped to enable flexible and generalizable visual task execution. This presentation will focus on the model elements that concern fixation change. Differing from most other models where a representation of image saliency drives selection of gaze patterns supporting a fixate-and-recognize vision process, here, saliency plays multiple roles all different from this common role. The first is stimulus-driven local feature conspicuity (stimulus-based attentional push), restricted to the visual periphery and enabling fixation changes overtly for reasons of surprise, novelty and exploration, based on the AIM framework (Bruce & Tsotsos 2009). Another is an object-centred conspicuity to drive central visual field fixation changes. These changes may be covert or short-range overt, and are intended to examine object components for purposes such as description, comparison or discrimination as well as pursuit. The central-peripheral distinction is imposed not only because of the retinal receptor anisotropy, but also to solve the boundary problem present in layered hierarchical representations (Tsotsos 2011). A third representation is task-specific attentional pull. This represents priority or urgency (conspicuity among unfinished components in the task domain) to attend a particular location, feature or object related to current task. Cognitive Programs lay out the temporal and causal sequence of operations comprising a task (Tsotsos & Kruijne 2014), working memory stores completed task elements, and the disparity at a given time between program and memory lead to task element priority. The latest results from the first stages of experimentation with STAR will round out the presentation.
Bruce, N.D.B., Tsotsos, J.K. (2009). Saliency, Attention, and Visual Search: An Information Theoretic Approach, J. of Vision, 9:3, p1-24.
Tsotsos, J.K. (2011). A Computational Perspective on Visual Attention, The MIT Press.
Tsotsos, J.K., Kruijne W. (2014).Cognitive programs: Software for attention's executive, Frontiers in Psychology: Cognition 5:1260.
John K. Tsotsos is Distinguished Research Professor of Vision Science at York University. He is Director of the Centre for Innovation in Computing at Lassonde, Canada Research Chair in Computational Vision, and is a Fellow of the Royal Society of Canada.
He received his doctorate in Computer Science from the University of Toronto. He did a postdoctoral fellowship in Cardiology at Toronto General Hospital and then joined the University of Toronto on faculty in both Computer Science and in Medicine, where he stayed for 20 years, 10 of which were as a Fellow of the Canadian Institute for Advanced Research. He then moved to York University serving as Director of the Centre for Vision Research for 7 years. Visiting positions were held at the University of Hamburg, Polytechnical University of Crete, Center for Advanced Studies at IBM Canada, INRIA Sophia-Antipolis, and the Massachusetts Institute of Technology.
He has published in computer science, neuroscience, psychology, robotics and bio-medicine. Current research has a main focus in developing a comprehensive theory of visual attention in humans. A practical outlet for this theory forms a second focus, embodying elements of the theory into the vision systems of mobile robots.