Botanical Tree Perceived Realism Metric
Abstract
In this thesis a novel metric for evaluating botanical plant perception is proposed. Firstly, utilizing diverse data sets of both synthetic and organic plants a unifying representation is found. This makes the presented approach generalizable to any type of tree model. Large data set of tree models from various sources both realistic and synthetic was assembled. Through a comparative perceptual study, every plant was sorted, and following an analysis labeled in terms of apparent realism. Since the specific parameters contributing to realistic perception are not fully understood, this research identifies key features based on observed data. Deep neural network and decision tree classifiers were trained and provided significant results on the data set and can predict perceived realism on new data as illustrated by a secondary study.
Degree
M.Sc.
Advisors
Benes, Purdue University.
Subject Area
Artificial intelligence|Plant sciences
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.