Advisor

Melanie Mitchell

Date of Award

9-28-2017

Document Type

Thesis

Degree Name

Master of Science (M.S.) in Computer Science

Department

Computer Science

Physical Description

1 online resource (vi, 48 pages)

DOI

10.15760/etd.5824

Abstract

For the last several years, convolutional neural network (CNN) based object detection systems have used a regression technique to predict improved object bounding boxes based on an initial proposal using low-level image features extracted from the CNN. In spite of its prevalence, there is little critical analysis of bounding-box regression or in-depth performance evaluation. This thesis surveys an array of techniques and parameter settings in order to further optimize bounding-box regression and provide guidance for its implementation. I refute a claim regarding training procedure, and demonstrate the effectiveness of using principal component analysis to handle unwieldy numbers of features produced by very deep CNNs.

Persistent Identifier

http://archives.pdx.edu/ds/psu/22151

Share

COinS