Biologically inspired object categorization in cluttered scenes

Show full item record

Title: Biologically inspired object categorization in cluttered scenes
Author: Peerasathien, Theparit
Abstract: The purpose of the thesis, Biologically Inspired Object Categorization system, is to provide an automatic system to classify the real-world images into categories. Generally, computer algorithms classify objects with much lower efficiency than human. Furthermore, some images with complex features such as cat and dog faces are difficult to be classified by ordinary computer algorithms. Therefore, the simulation of the structure and process of a mammalian’s visual cortex is created, which functions similarly to a human’s visual cortex, by using a computer. In this paper, I am presenting a biologically inspired neural network system which processes the images in a hierarchical order, starting from emulation of the retina cells to the virtual cortex. The goal of the network is to recognize objects in images which serve to answer the “what” objects that are in the scene. “What” is one of the pathways the brain recognizes of an object, aside from the ‘where’ pathway. The system can be used in many applications such as categorizing cat and dog faces individually or clustering automobiles in an urban scene.
Record URI: http://hdl.handle.net/1850/7930
Date: 2008

Files in this item

Files Size Format View
TPeerasathienThesis04-2008.pdf 5.336Mb PDF View/Open

The following license files are associated with this item:

This item appears in the following Collection(s)

Show full item record

Search RIT DML


Advanced Search

Browse