NBNN-DF: NBNN WITH DISCRIMINATIVE FEATURES
KAIST Tech. Report (CS-TR-2012-371)
(Submitted to ICME 2013)
|
This figure shows the overall flow of NBNN-DF.
|
Abstract
We propose a novel, Na¡§©¥ve Bayes Nearest Neighbor (NBNN)
classifier considering Discriminative Features, NBNN-DF,
for improving both classification accuracy and query-time
performance. Unlike the original NBNN method, we define
discriminative features among all the descriptors extracted
from training and query images, and perform NBNN with
those discriminative features. To define discriminative features
we measure a discriminative power for each feature
based on a ratio of posterior probability that the feature is
located in its positive class to that in its negative class. While
it is easy to measure discriminative power for features extracted
from training images, we face the chicken-and-egg
problem for a query image, whose class type is unknown. To
address this problem we hypothesize potential class types of
the query image and perform NBNN with discriminative features
under its hypothetical classes, while considering a confidence
level of each hypothesis. We have tested our method on
the Caltech101 dataset, and compared it against other stateof-
the-art techniques. We found that our method, NBNNDF,
achieves up to 34% relative accuracy improvement over
prior techniques. Our technique achieves this improvement,
while improving the overall query-time performance by using
a smaller number of features.
Contents
Paper:
NBNN-DF: NBNN WITH DISCRIMINATIVE FEATURES,
KAIST Tech. Report (CS-TR-2012-371), 2012 (Submitted to ICME 2013)
Related Links
SGLab
Dept. of Computer Science
KAIST
373-1 Guseong-dong, Yuseong-gu, Daejeon, 305-701
South Korea
sglabkaist dot gmail dot com