Neural network for dynamic binding with graph representation: form, linking, and depth-from-occlusion.

J R Williamson

doi:10.1162/neco.1996.8.6.1203

Home/Archive/PubMed

PubMed · 8768392

Neural network for dynamic binding with graph representation: form, linking, and depth-from-occlusion.

J R Williamson

View original source ↗Export metadata ↓

Abstract

A neural network is presented that explicitly represents form attributes and relations between them, thus solving the binding problem without temporal coding. Rather, the network create a graph representation by dynamically allocating nodes to code local form attributes and establishing arcs to link them. With this representation, the network selectively groups and segments in depth objects based on line junction information, producing results consistent with those of several recent visual search experiments. In addition to depth-from-occlusion, the network provides a sufficient framework for local line-labeling processes to recover other three-dimensional (3-D) variables, such as edge/surface contiguity, edge slant, and edge convexity.

Explore related subjects

Form Perception Neural Networks, Computer Photic Stimulation Journal Article Research Support, U.S. Gov't, Non-P.H.S.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

J R Williamson. 1996-08-15. Neural network for dynamic binding with graph representation: form, linking, and depth-from-occlusion.. https://doi.org/10.1162/neco.1996.8.6.1203

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2007

Absence of flash-lag when judging global shape from local positions.

When a flash is presented aligned with a moving stimulus, the former is perceived to lag behind the latter (the flash-lag effect). We study whether this mislocalization occurs when a positional judgment is not required, but a veridical spatial relationship between moving and flashed stimuli is needed to perceive a global shape. To do this, we used Glass patterns that are formed by pairs of correlated dots. One dot of each pair was presented moving and, at a given moment, the other dot of each pair was flashed in order to build the Glass pattern. If a flash-lag effect occurs between each pair of dots, we expect the best perception of the global shape to occur when the flashed dots are presented before the moving dots arrive at the position that physically builds the Glass pattern. Contrary to this, we found that the best detection of Glass patterns occurred for the situation of physical alignment. This result is not consistent with a low-level contribution to the flash-lag effect.

Form Perception↗

PubMed2006

Object recognition and segmentation by a fragment-based hierarchy.

How do we learn to recognize visual categories, such as dogs and cats? Somehow, the brain uses limited variable examples to extract the essential characteristics of new visual categories. Here, I describe an approach to category learning and recognition that is based on recent computational advances. In this approach, objects are represented by a hierarchy of fragments that are extracted during learning from observed examples. The fragments are class-specific features and are selected to deliver a high amount of information for categorization. The same fragments hierarchy is then used for general categorization, individual object recognition and object-parts identification. Recognition is also combined with object segmentation, using stored fragments, to provide a top-down process that delineates object boundaries in complex cluttered scenes. The approach is computationally effective and provides a possible framework for categorization, recognition and segmentation in human vision.

Form Perception↗

PubMed2006

Dynamics of shape interaction in human vision.

Spatial context can alter perceived shape, and temporal context can influence the perception of a stimulus. We sought to determine the time course of shape interactions by using a paradigm in which closed shape contours are laterally displaced over space and time. Target and masks are separated by various stimulus onset asynchrony (SOA) values, yielding forward, backward, and simultaneous masking conditions. Results indicate that spatial lateral interactions of shape are amplified by temporal asynchrony, reaching a peak at SOAs of 80-110 ms. Mask amplitude scales all effects and masking is shape specific. When a single mask follows the target, both spatial configuration and mask onset transient are critical in determining depth of masking. When the target is followed by two sequential masks, the possibility of apparent motion determines whether one or both masks drive masking. These findings suggest that temporal interactions of shape are dependent on an interactive combination of shape specificity and transients, that apparent motion plays a modulatory role, and that target shape is determined after a temporal window, not at its onset.

Form Perception↗