Skip to main content.
Advanced search >
<< Back to previous page Print

<< Monday, September 17, 2018 >>

Remind me

Tell a friend

Add to my Google calendar (bCal)

Download to my calendar

Bookmark and ShareShare

​Graduate Students Seminar

Seminar: Oxyopia Seminar | September 17 | 11:10 a.m.-12:30 p.m. | 489 Minor Hall

Vivek Labhishetty, PhD; Baladitya Yellapragada, PhD

Neuroscience Institute, Helen Wills

Vivek Labhishetty's Abstract
Retinal-Conjugate Surfaces: The Blur Horopter
When we fixate at an object, the image of that object is brought to sharp focus on the fovea due to the eye’s accommodation. Other objects in the periphery may be farther or nearer than best focus on those parts of the retina. We measured the shape of surface of best focus in the world as the eye accommodates to different distances. To do this, we used wavefront-aberration data from 15 emmetropic eyes in the central 30o of the visual field. The resulting surface—the retinal conjugate—is consistently pitched top-back and rotated slightly nasal-back. We show that those effects are consistent with the statistics of the natural environment for people engaged in everyday tasks. The surface is also reasonably consistent with the empirical horopter, the position in the world where stereopsis is most precise. The retinal-conjugate surface and horopter conform to habitual environmental statistics.

Baladitya Yellapragada's Abstract
Motion Selectivity of Neurons in Self-Driving Networks
We investigated if optical flow filters were implicitly learned by a neural network trained to drive a vehicle. The network was not trained to predict optical flow across the frames, but, through a series of controlled experiments, we claim that optical flow filters are present in the network. However, this appears to be only the case for sideways flows more relevant for steering predictions. For motor throttle predictions, the network looks at the variance of the pixels over time rather than computing optical flow. In addition, the filters that are likely used for motor throttle predictions dominate primarily in the middle of the network.