國立虎尾科技大學 |

Image Classification with Hierarchical Datasets.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Image Classification with Hierarchical Datasets./
Author:	Wadkar, Abhijeet Ramesh.
Description:	1 online resource (57 pages)
Notes:	Source: Masters Abstracts International, Volume: 56-06.
Contained By:	Masters Abstracts International56-06(E).
Subject:	Computer science. -
Online resource:	click for full text (PQDT)
ISBN:	9780355135480

Image Classification with Hierarchical Datasets.
Wadkar, Abhijeet Ramesh.

Image Classification with Hierarchical Datasets. - 1 online resource (57 pages)

Source: Masters Abstracts International, Volume: 56-06.

Thesis (M.S.)

Includes bibliographical references

This item is not available from ProQuest Dissertations & Theses.

The image classification problem has become exceedingly important with applications in fields such as self-driving cars and robotics. In this research, we use the hierarchy of object categories to improve state-of-the-art convolutional neural network (CNN) image classification techniques. We use semantic hierarchies in the labels of images to integrate prior knowledge about the real-world hierarchies of concepts or objects while training our CNN models. We start with an architecture based on the AlexNet and present a hierarchy aware loss function for training the CNNs and evaluate the top-1 and top-5 predictions of our models with respect to a baseline model on the Caltech-256 and ImageNet datasets. The results of our best architecture show an increase of 6.16% in top-1 and 4.54% increase in top-5 accuracy for the Caltech-256 dataset as well as an average increase of 4.09% in top-1 and 1.83% in top-5 accuracy in the ImageNet dataset over the baseline. Finally, we conduct an extensive analysis of our models' performance and showcase that our CNN architecture is able to classify better in generalized concepts higher up in the semantic hierarchy.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2018

Mode of access: World Wide Web

ISBN: 9780355135480Subjects--Topical Terms:

573171
Computer science.
Index Terms--Genre/Form:

554714
Electronic books.

Image Classification with Hierarchical Datasets.
LDR:02501ntm a2200373Ki 4500 001 910748
005 20180517112609.5
006 m o u
007 cr mn||||a|a||
008 190606s2017 xx obm 000 0 eng d
020 $a 9780355135480
035 $a (MiAaPQ)AAI10274722
035 $a (MiAaPQ)umbc:11621
035 $a AAI10274722
040 $a MiAaPQ $b eng $c MiAaPQ
099 $a TUL $f hyy $c available through World Wide Web
100 1 $a Wadkar, Abhijeet Ramesh. $3 1182196
245 1 0 $a Image Classification with Hierarchical Datasets.
264 0 $c 2017
300 $a 1 online resource (57 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 56-06.
500 $a Adviser: Tim Oates.
502 $a Thesis (M.S.) $c University of Maryland, Baltimore County $d 2017.
504 $a Includes bibliographical references
506 $a This item is not available from ProQuest Dissertations & Theses.
520 $a The image classification problem has become exceedingly important with applications in fields such as self-driving cars and robotics. In this research, we use the hierarchy of object categories to improve state-of-the-art convolutional neural network (CNN) image classification techniques. We use semantic hierarchies in the labels of images to integrate prior knowledge about the real-world hierarchies of concepts or objects while training our CNN models. We start with an architecture based on the AlexNet and present a hierarchy aware loss function for training the CNNs and evaluate the top-1 and top-5 predictions of our models with respect to a baseline model on the Caltech-256 and ImageNet datasets. The results of our best architecture show an increase of 6.16% in top-1 and 4.54% increase in top-5 accuracy for the Caltech-256 dataset as well as an average increase of 4.09% in top-1 and 1.83% in top-5 accuracy in the ImageNet dataset over the baseline. Finally, we conduct an extensive analysis of our models' performance and showcase that our CNN architecture is able to classify better in generalized concepts higher up in the semantic hierarchy.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2018
538 $a Mode of access: World Wide Web
650 4 $a Computer science. $3 573171
650 4 $a Information science. $3 561178
650 4 $a Artificial intelligence. $3 559380
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0723
690 $a 0800
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a University of Maryland, Baltimore County. $b Computer Science. $3 1179407
773 0 $t Masters Abstracts International $g 56-06(E).
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10274722 $z click for full text (PQDT)