Visual Perception of Depth from Occlusion: A Neural Network Model
Abstract
During this period, we have made substantial progress in understanding how objects are discriminated by the visual system. A number of papers describing our results are enclosed. Major goals of the project have been accomplished in several areas: (1) Completion of a beta release version of a novel neural network simulator, NEXUS; (2) Development of a parallel version of the NEXUS simulator which allows anatomically interconnected networks to be simultaneously simulated on different workstations linked by ethernet connections; (3) Continuing research using our model of how the visual system extracts depth-from-occlusion. Considerable progress with regard to how surfaces may be represented. Numerous simulations of responses to both real and illusory objects; and (4) Development of related cortically-based models of color visual processing and texture discrimination. Color and texture are used to detect contours in images, and these contours can then be used by our depth-from-occlusion model to determine the relative depth of the colored or textured surfaces.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA249035
Entities
People
- Leif H. Findel
Organizations
- University of Pennsylvania