Keynotes

Ryad Benosman - What is neuromorphic computer vision? Sensors, theory and applications.
Vision and natural computation, Institut de la Vision, Paris, France

Abstract. In this presentation, I will present neuromorphic, event-based approaches to the detection and processing of visual information and new approaches to learning based solely on the notion of a precise timing. Conventional imaging sensors suffer from severe limitations imposed by their operating principle based on stroboscopic image acquisition. These sensors acquire visual information in the form of a series of "snapshots" recorded at discrete temporal instants, thus quantified over time at a predetermined rate, resulting in a limited temporal resolution, a low dynamic range and a high degree of redundancy in the acquired data. Nature suggests a different approach: biological vision systems are controlled and monitored by events occurring in the scene, and not - like conventional image sensors - by artificially created synchronization and control signals that have no connection to the source of visual information. Translating the frameless paradigm of biological vision into artificial imaging systems implies that control of visual information acquisition is no longer imposed externally on a pixel matrix, but that decision making is transferred to each pixel, which processes its own information individually. I will present the fundamental principles underlying these approaches to the detection and processing of asynchronous visual information inspired by biological events and biology, and showing the strengths and weaknesses of this type of systems. I will show that bio-inspired vision systems have the potential to replace conventional image-based vision processing and acquisition systems and set new standards in terms of data compression, dynamic range, time resolution and energy efficiency in applications such as 3D vision. object tracking, motor control, visual feedback loops and real-time automatic learning at several hundred kHz.

______

Georgia Gregoriou - The role of executive control areas in attentional selection. Insights from electrophysiology
Laboratory of Visual Cognition, Medical School, University of Crete, Greece

Abstract. Converging evidence from electrophysiological recordings and lesion studies in non-human primates (NHPs) has revealed that in NHPs, as in humans, the prefrontal (PFC) and parietal (PPC) cortices are critical for the control of selective attention. It is widely held that PFC and PPC provide top-down inputs to the visual cortex to selectively enhance the representation of behaviorally relevant stimuli. This talk will discuss recent data on the role of PFC and PPC in guiding attention and will suggest mechanisms through which executive control areas modulate processing in visual cortex with attention. Based on recent results from our lab, I will compare activity in PFC and PPC during spatial and feature-based attention and will highlight the distinct contributions of the two areas in attentional processes. Furthermore, I will show work that aims to identify critical PFC sources of attentional feedback to mid-level visual cortical areas.

______

Eli Brenner - The visual control of goal-directed arm movements
Faculty of Human Movement Sciences, Vrije Universiteit, Amsterdam, The Netherlands

Abstract. I will talk about how visual information about a target’s position and motion is used to guide one’s hand to the target. I will present several experiments that examined how movements are influenced by small perturbations and by sequential regularities. I will present evidence that people continuously update their movements on the basis of the latest available information about the target’s position. People appear to consider the target’s velocity across extended periods of time. They do not consider acceleration at all. They get away with using such incomplete information by compensating for recent errors. I will argue that precise interception is achieved using rather limited but constantly updated visual information.