August 2016
Volume 16, Issue 12
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2016
Visual search in natural scenes: Normative modeling of the target absent case
Author Affiliations
  • Jared Abrams
    Center for Perceptual Systems, University of Texas at Austin
  • Wilson Geisler
    Center for Perceptual Systems, University of Texas at Austin
Journal of Vision September 2016, Vol.16, 1160. doi:10.1167/16.12.1160
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Jared Abrams, Wilson Geisler; Visual search in natural scenes: Normative modeling of the target absent case. Journal of Vision 2016;16(12):1160. doi: 10.1167/16.12.1160.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Normative models of search provide a benchmark for evaluating human performance and identifying rational behavioral strategies. Previous work has identified the entropy limit minimization (ELM) observer as an optimal fixation selection model.1 The ELM observer selects fixation locations that maximally reduce uncertainty about the location of the target. We compare this with a maximum a posteriori (MAP) model, which selects fixation locations that have the greatest posterior probability of containing the target. Previously, these models were specified for the case where the target is always present. In natural tasks, a target may not be present in the scene. Here, we extend these models to produce predictions for the case where the target may be present or absent. Human observers participated in a search experiment where a target was present in a natural scene on half of the trials. Observers reported the location of the target or indicated its absence. Our observers required 2.25 times as many fixations to indicate target absent (9 fixations) versus target present (4 fixations) while maintaining a high degree of accuracy. Both the ELM and MAP models are faster and more accurate than human searchers. We scaled the single fixation map of detectability (d' map) in each model to match the hit rate (91.0%), correct rejection rate (90.8%), and median number of fixations (6.5) of the observers. We found that the ratio of target absent fixations to target present fixations was 2.5 for the MAP, compared with a value of 2.27 for the ELM. The similarity in the ratios of target absent fixations to target present fixations between the human and ELM observers suggests that the ELM describes human fixations in natural scenes better than the MAP. 1Najemnik, J. & Geisler W.S. (2009) Vis. Res., 49, 1286-1294.

Meeting abstract presented at VSS 2016

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×