Evaluating Human Visual Search Performance by Monte Carlo methods and Heuristic model Giacomo Veneri , Elena Pretegiani, Pamela Federighi, Francesca Rosini, Antonio Federico & Alessandra Rufa University of Siena Department of Nerurology & Neurosurgery and Behavioral Sciences
Evaluating Human Visual Search Performance by Monte Carlo methods and Heuristic model
May 25, 2015
Visual search is an everyday activity that enables
humans to explore the real world. Given the visual input,
during a visual search, it’s required to select some aspects of the input in order to move to the next location. Exploration is guided by two factors: saliency of image (bottom-up) and endogenous mechanism (top-down). These two mechanisms interact to perform an efficient visual search. We developed a stochastic model, the ”break away from fixations” (BAF), to emulate the visual search on a high cognitively demanding task such as a trail making test (TMT). The paper reports a case study providing evidence that human exploration performs an efficient visual search based also on an internal model of regions already explored.
humans to explore the real world. Given the visual input,
during a visual search, it’s required to select some aspects of the input in order to move to the next location. Exploration is guided by two factors: saliency of image (bottom-up) and endogenous mechanism (top-down). These two mechanisms interact to perform an efficient visual search. We developed a stochastic model, the ”break away from fixations” (BAF), to emulate the visual search on a high cognitively demanding task such as a trail making test (TMT). The paper reports a case study providing evidence that human exploration performs an efficient visual search based also on an internal model of regions already explored.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Evaluating Human Visual Search Performance by
Monte Carlomethods and Heuristic model
Giacomo Veneri, Elena Pretegiani, Pamela Federighi, Francesca Rosini, Antonio Federico & Alessandra Rufa
University of SienaDepartment of Nerurology & Neurosurgery and Behavioral
Sciences
Human Visual System The main mechanisms of Human Brain
Eye Tracking Method How to register Eye movements
How to model Visual Search The Model Results of healthy subjects Results of patients Conclusions
Summary
60% of Brain is
dedicated
to Vision
Human Visual System
Evaluating Human Visual Search Performance by Monte Carlo
methods and Heuristic model
Saccade aprox 600deg/secSaccade aprox 600deg/sec
Exp1/5
Human is foveate animalHuman is foveate animal
E
E
E
EE
E
EESpot AttentionSpot Attention
Exp2/5
SaliencySaliency
Exp4/5
The optic radiations, one on each side of the brain, carry information from the thalamic lateral geniculate nucleus to layer 4 of the visual cortex.
The lateral geniculate nucleus (LGN) is a sensory relay nucleus in the thalamus of the brain. The LGN consists of six layers in humans and other primates starting from catarhinians, including cercopithecidae and apes.
The visual cortex is the most massive system in the human brain and is responsible for processing the visual image.
Human Visual System pathway
Eye Movement
fixation, in which the eyes are directed toward a motionless object
saccades, in which the eyes move very rapidly from one location to another
Other smooth pursuit, in which the eyes
move steadily to track a moving object;
vergence, in which the eyes move simultaneously in opposite directions to obtain or maintain single binocular vision.
Human Visual System
Where: where is the object? What: which object to see?
The overall system: Eye Movement
Saccade, vergence, fixation Attention Memory/Working memory Object
segmentation/identification Peripheral vision Saliency
Where
What
Visual Serach is Pseudo Random
Visual processing pathways in monkeys. Areas in the dorsal stream, having primarily visuospatial functions, are shown in green, and areas in the ventral stream, having primary object recognition functions, are shown in red. Lines connecting the areas indicate known anatomical connections, with heavy arrowheads indicating feed-forward connections from lower-order areas to higher-order ones and open arrowheads indicating feedback connections from higher-order areas to lower-order ones. Solid lines indicate connections from both central and peripheral visual field representations, and dotted lines indicate connections restricted to peripheral field representations. Shaded region on lateral view of the brain indicates extent of cortex included in the diagram. (From ref. 85; for further details of the visual areas, see ref. 86.)
Monkey Visual System
Ungerleider L G et al. PNAS 1998;95:883-890
Where
What
Main Block
Saliency of Image/scene Peripheral vision
AttentionInternal statusWorking memory
The key idea
parameters
subjects
patients
Subject break away from latest fixation
The temporal window is about 1000ms (3 fixations)
[Veneri & Pretegiani 2010]
The role of latest fixations
12/04/23Giacomo Veneri - EVALab 15
Periphera Vision
The probability to reach the target when the fixation is far 4degree
[Findlay 2000]
Results
Evaluating Human Visual Search Performance by Monte Carlo
methods and Heuristic model
Peripheral Vision Inhibited
Normal subjects
Visual System Optimization
Conclusion
Evaluating Human Visual Search Performance by Monte Carlo
methods and Heuristic model
Conclusion
12/04/23 Giacomo Veneri - EVALab 23
The model is able to emulate human visuals search exploration.
Human optimizes the exploration in order to reduce eye movements and energy consumptions.
The model explains the exploration of subjects
Evaluating Human Visual Search Performance by
Monte Carlomethods and Heuristic model
Giacomo Veneri, Elena Pretegiani, Pamela Federighi, Francesca Rosini, Antonio Federico & Alessandra Rufa
Thank youThank you
Related Documents
