סמינר מחלקתי

A simulation model of computer-aided visual search

Herman Ravkin – M.Sc. student

Abstract:

We perform visual search whenever we look for a specific object in our environment.  In recent years visual search in areas, such as medicine or military applications is supported by automated detection systems. The question arises how computer aided object detection interacts with human decision making. The purpose of this study is to create a simulation model of the human decision process in a visual search task aided by computer, and to acquire a deeper understanding of cue influences on this decision process. The model addresses the effect of the cue: its influence on the visual scanning and its influence on items classification.

The comparison of the results of the simulation model to empirical data shows that our simulation model can explain a significant part of the process variability, especially with regard to the decision itself. From the analysis of the simulation results we found that cues might affect the rate of decision convergence in humans and the number of evidences they are willing to collect before making the decision. This fact can explain why computer-aided decision quality is worse than would be expected from an information-theoretic point of view: humans look for a "good enough" decision, and they are willing to trade a marginal improvement in decision quality for faster processes. We also found the secondary effect of cues through the change in humans' scanning strategy.

This work was performed under the supervision of Prof. Joachim Meyer.

ההרצאה תתקיים ביום חמישי 21.5.15, בשעה 14:00, בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.