國立虎尾科技大學 |

Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance./
Author:	Naha, Shujon.
Description:	1 online resource (108 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-10, Section: B.
Contained By:	Dissertations Abstracts International85-10B.
Subject:	Computer science. -
Online resource:	click for full text (PQDT)
ISBN:	9798381973440

Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance.
Naha, Shujon.

Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance. - 1 online resource (108 pages)

Source: Dissertations Abstracts International, Volume: 85-10, Section: B.

Thesis (Ph.D.)--Indiana University, 2024.

Includes bibliographical references

This thesis discusses the significance of generalization to novel objects in computer vision and focuses particularly on utilizing semantic guidance for achieving such generalizations. In computer vision, generalization involves training a model on a specific set of known objects and expecting it to perform well on unseen or rarely seen objects. Here, we show that semantic guidance can be used as a highly effective tool to infer about novel objects. In particular, we study three different computer vision problems focusing on generalizing to novel objects using semantic guidance. In the first problem, we localize novel attended objects from egocentric views. We show that the concept of 'human attention' can be utilized as semantic guidance and used to localize attended objects which were never seen before. In the second problem, we explore the fine-grained object part segmentation task, where the target object is absent from the training data. We show that we can use keypoints as semantic guidance to transfer part knowledge from seen to unseen objects with similar semantic body parts. We proved the effectiveness of the approach by segmenting the body parts of unseen quadruped animals. Finally, in the third problem, we study the problem of novel object 6DOF pose estimation using 3D hand pose as the semantic guidance in a few-shot manner. Here we show that the hands can be used as a strong contextual cue to understand 3D pose of novel hand manipulated objects without even using any training at all.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381973440Subjects--Topical Terms:

573171
Computer science.
Subjects--Index Terms:

Computer visionIndex Terms--Genre/Form:

554714
Electronic books.

Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance.
LDR:02910ntm a22003977 4500 001 1149693
005 20241022112616.5
006 m o d
007 cr bn ---uuuuu
008 250605s2024 xx obm 000 0 eng d
020 $a 9798381973440
035 $a (MiAaPQ)AAI30996838
035 $a AAI30996838
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Naha, Shujon. $3 1476018
245 1 0 $a Three Studies on Generalizing to Novel Objects in Computer Vision Using Semantic Guidance.
264 0 $c 2024
300 $a 1 online resource (108 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: Dissertations Abstracts International, Volume: 85-10, Section: B.
500 $a Advisor: Crandall, David J.
502 $a Thesis (Ph.D.)--Indiana University, 2024.
504 $a Includes bibliographical references
520 $a This thesis discusses the significance of generalization to novel objects in computer vision and focuses particularly on utilizing semantic guidance for achieving such generalizations. In computer vision, generalization involves training a model on a specific set of known objects and expecting it to perform well on unseen or rarely seen objects. Here, we show that semantic guidance can be used as a highly effective tool to infer about novel objects. In particular, we study three different computer vision problems focusing on generalizing to novel objects using semantic guidance. In the first problem, we localize novel attended objects from egocentric views. We show that the concept of 'human attention' can be utilized as semantic guidance and used to localize attended objects which were never seen before. In the second problem, we explore the fine-grained object part segmentation task, where the target object is absent from the training data. We show that we can use keypoints as semantic guidance to transfer part knowledge from seen to unseen objects with similar semantic body parts. We proved the effectiveness of the approach by segmenting the body parts of unseen quadruped animals. Finally, in the third problem, we study the problem of novel object 6DOF pose estimation using 3D hand pose as the semantic guidance in a few-shot manner. Here we show that the hands can be used as a strong contextual cue to understand 3D pose of novel hand manipulated objects without even using any training at all.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer science. $3 573171
650 4 $a Computer engineering. $3 569006
653 $a Computer vision
653 $a Deep learning
653 $a Domain generalization
653 $a Machine learning
653 $a Zero-shot learning
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0464
690 $a 0800
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Indiana University. $b Computer Science. $3 1476019
773 0 $t Dissertations Abstracts International $g 85-10B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30996838 $z click for full text (PQDT)