Statistical Correlations Between 3D and 2D Scenes
We are interested in discovering the statistical regularities in natural 3D scenes. This involves studying the correlations between range images and optical images, as well as studying statistical regularities in neurons' responses to natural scene images.
|
Neural Representation of 3D Surface Structure
How does the brain represent and compute 3D surfaces? Our work is based on the idea that surface inference is an active process that involves the interpretation of multiple intrinsic factors and the reconciliation of these factors.
|
Motion Estimation via Bayesian Belief Propagation
Little is known about the computational mechanisms for motion computation in the brain. We are evaluating the role of recurrent feedback on motion velocity estimation in a hierarchical framework by studying the problem of motion velocity estimation in ambiguous situations.
|
Computational Methods for Object Recognition in Natural Scenes
Like motion and surface perception, recognition of the identities of objects is also highly influenced by contextual information. We are studying the manner in which predictions based on contextual cues
can improve the accuracy and speed of object segmentation, tracking, and
recognition.
|
Time Course of Learning Perceptual Pop-out in V1 and V2
This research focuses on determining the neural substrate of changes
in perception and behavior after sensory experience, also known as
perceptual learning. Our primary goal is to
explore the neural substrate of perceptual learning to determine how
neuronal changes filter through the visual hierarchy.
|
Dynamical Adaptation and Gain Control
The study of the effect of different stimuli on neuron properties at different time scales, from the order of seconds to weeks. In addition, we study the changes on a very long time scale (years) due to the environmental impact. What are the computational principles underlying such changes?
|
Decoding Neuronal Activities in V1
We are striving to go beyond the classical neurophysiological tuning curves in evaluating the computation of neurons. We approach this by attempting to accurately encode and decode neuronal activities in awake monkeys using Volterra kernels and particle filtering.
|
