Research Interests
Accurately estimating physical variables from sensory signals (images) does not ensure an agent’s success in real-world tasks. Visual perception needs to be: (i) robust against numerous contextual variables originating from many independent physical processes in nature; (ii) flexible to adapt the estimation process to accommodate the agent’s internal and external needs and specific environments; and (iii) able to dynamically prioritize certain variables constant presence of risks and opportunities in the world. Integrating these three capacities into a single system (perceptual agent) imposes a strong constraint on the visual system.
These capabilities are demonstrated by humans’ remarkable ability to perform various simple visual tasks, exhibiting rich, diverse, and sometimes odd behavior. I aim to progressively build image-computable pseudo-agent models capable of predicting outcomes for various simple experiments to improve our understanding of perceptual agency. Thus, I am interested in constructing datasets that will showcase these capabilities to develop standard models of visual perception by benchmarking various pseudo-agent models against the datasets.
Active Projects
Confidence in Global Motion Direction Discrimination
Estimation of the direction of visual motion is essential for most living beings. It is also vital to accurately judge the confidence about the estimated direction of motion. In the formation of confidence, both the visual reliability of the stimuli and other visual-cognitive factors play a role. For a straightforward global motion direction discrimination task, we aim to identify the role of these mechanisms by measuring and modeling the human performance (perceptual and metacognitive).
Visual Target Detection Under Uncertainty
Detection of visual targets is integral to survival and everyday functioning. In the real world, the visual system operates under very high levels of extrinsic uncertainty (multiple simultaneous dimensions of uncertainty) about target and background properties. The aim of the project is to better understand how the human visual system detects objects while being uncertain about the properties of the object and its background (i.e., being uncertain about the exact visual pattern to look for).
Oluk, C., & Geisler, W. S. (2022). Effects of Target-Amplitude and Background-Contrast Uncertainty Predicted by a Normalized Template-Matching Observer. Journal of Vision, 23(12), 8. GitHub Page.
My Dissertation is available online.
Previous Projects
Visual Perception of 3D Slant
Binocular stereo cues are important for discriminating 3D surface orientation, especially at near distances. We devised a single-interval task where observers discriminated the slant of a densely textured planar test surface relative to a textured planar surround reference surface. Slant discrimination performance was measured as a function of the reference slant and the level of uncorrelated white noise added to the test-plane images in the left and right eyes. We compared human performance with an approximate ideal observer and two subideal observers.
Oluk, C., Bonnen, K., Burge, J., Cormack, L. K., & Geisler, W. S. (2022). Stereo slant discrimination of planar 3D surfaces: Frontoparallel versus planar matching. Journal of Vision, 22(5), 6-6. GitHub Page.
As part of this project, we also derived the ideal observers, where the task is to estimate the absolute disparity when the IID texture is a different unknown random sample on each trial, and where “internal noise” is represented by adding some level of independent Gaussian pixel noise that is uncorrelated in the left and right images.
Oluk, C., & Geisler, W. S. (2020). Ideal Observers for the estimation of disparity in random-pixel stereograms. Journal of Vision, 20(11), 578-578. GitHub Page.