A model of active visual search with object-based attention guiding scan paths

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Attention

Background:

• Monkey visual search behavior demonstrates a preference for color-guided scan paths over orientation-guided ones when searching for conjunction targets.
• Previous models have struggled to replicate the temporal dynamics of attention in visual search tasks.

Purpose of the Study:

• To develop and validate an active vision model that can replicate monkey visual search behavior, specifically the guidance of scan paths by feature conjunctions.
• To investigate the neural mechanisms underlying the development of object-based attention and its role in guiding behavior.

Main Methods:

• An active vision model employing biased competition was developed to simulate visual search.
• The model's neural responses and scan path generation were analyzed and compared to psychophysical data and single-cell recordings from monkeys.

Main Results:

• The model successfully replicated the observed preference for color-guided scan paths.
• Attention in the model evolved from an early spatial effect to a later object-based effect, mirroring experimental findings.
• The model demonstrated temporal precision in reproducing these attentional effects at a systems level.

Conclusions:

• The biased competition model provides a viable explanation for how object-based attention develops and guides behavior in visual search.
• This model offers a systems-level replication of psychophysical scan paths, advancing our understanding of visual attention mechanisms.