Group optimization for multi-attribute visual embedding
Understanding semantic similarity among images is the core of a wide range of computer graphics and computer vision applications. However, the visual context of images is often ambiguous as images that can be perceived with emphasis on different attributes. In this paper, we present a method for lea...
Main Authors: | , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2018-09-01
|
Series: | Visual Informatics |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2468502X18300408 |
id |
doaj-8752094740af422b80fdc9e9326bad29 |
---|---|
record_format |
Article |
spelling |
doaj-8752094740af422b80fdc9e9326bad292020-11-25T00:36:14ZengElsevierVisual Informatics2468-502X2018-09-0123181189Group optimization for multi-attribute visual embeddingQiong Zeng0Wenzheng Chen1Zhuo Han2Mingyi Shi3Yanir Kleiman4Daniel Cohen-Or5Baoquan Chen6Yangyan Li7School of Computer Science & Technology, Shandong University, Qingdao, China; Corresponding author.Department of Computer Science, University of Toronto, Toronto, CanadaViterbi School of Engineering, University of Southern California, CA, United StatesSchool of Computer Science & Technology, Shandong University, Qingdao, ChinaDNEG Visual Effects, London, United KingdomSchool of Computer Science, Tel Aviv University, Tel Aviv, IsraelSchool of Computer Science & Technology, Shandong University, Qingdao, ChinaSchool of Computer Science & Technology, Shandong University, Qingdao, ChinaUnderstanding semantic similarity among images is the core of a wide range of computer graphics and computer vision applications. However, the visual context of images is often ambiguous as images that can be perceived with emphasis on different attributes. In this paper, we present a method for learning the semantic visual similarity among images, inferring their latent attributes and embedding them into multi-spaces corresponding to each latent attribute. We consider the multi-embedding problem as an optimization function that evaluates the embedded distances with respect to qualitative crowdsourced clusterings. The key idea of our approach is to collect and embed qualitative pairwise tuples that share the same attributes in clusters. To ensure similarity attribute sharing among multiple measures, image classification clusters are presented to, and solved by users. The collected image clusters are then converted into groups of tuples, which are fed into our group optimization algorithm that jointly infers the attribute similarity and multi-attribute embedding. Our multi-attribute embedding allows retrieving similar objects in different attribute spaces. Experimental results show that our approach outperforms state-of-the-art multi-embedding approaches on various datasets, and demonstrate the usage of the multi-attribute embedding in image retrieval application. Keywords: Embedding, Semantic similarity, Visual retrievalhttp://www.sciencedirect.com/science/article/pii/S2468502X18300408 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Qiong Zeng Wenzheng Chen Zhuo Han Mingyi Shi Yanir Kleiman Daniel Cohen-Or Baoquan Chen Yangyan Li |
spellingShingle |
Qiong Zeng Wenzheng Chen Zhuo Han Mingyi Shi Yanir Kleiman Daniel Cohen-Or Baoquan Chen Yangyan Li Group optimization for multi-attribute visual embedding Visual Informatics |
author_facet |
Qiong Zeng Wenzheng Chen Zhuo Han Mingyi Shi Yanir Kleiman Daniel Cohen-Or Baoquan Chen Yangyan Li |
author_sort |
Qiong Zeng |
title |
Group optimization for multi-attribute visual embedding |
title_short |
Group optimization for multi-attribute visual embedding |
title_full |
Group optimization for multi-attribute visual embedding |
title_fullStr |
Group optimization for multi-attribute visual embedding |
title_full_unstemmed |
Group optimization for multi-attribute visual embedding |
title_sort |
group optimization for multi-attribute visual embedding |
publisher |
Elsevier |
series |
Visual Informatics |
issn |
2468-502X |
publishDate |
2018-09-01 |
description |
Understanding semantic similarity among images is the core of a wide range of computer graphics and computer vision applications. However, the visual context of images is often ambiguous as images that can be perceived with emphasis on different attributes. In this paper, we present a method for learning the semantic visual similarity among images, inferring their latent attributes and embedding them into multi-spaces corresponding to each latent attribute. We consider the multi-embedding problem as an optimization function that evaluates the embedded distances with respect to qualitative crowdsourced clusterings. The key idea of our approach is to collect and embed qualitative pairwise tuples that share the same attributes in clusters. To ensure similarity attribute sharing among multiple measures, image classification clusters are presented to, and solved by users. The collected image clusters are then converted into groups of tuples, which are fed into our group optimization algorithm that jointly infers the attribute similarity and multi-attribute embedding. Our multi-attribute embedding allows retrieving similar objects in different attribute spaces. Experimental results show that our approach outperforms state-of-the-art multi-embedding approaches on various datasets, and demonstrate the usage of the multi-attribute embedding in image retrieval application. Keywords: Embedding, Semantic similarity, Visual retrieval |
url |
http://www.sciencedirect.com/science/article/pii/S2468502X18300408 |
work_keys_str_mv |
AT qiongzeng groupoptimizationformultiattributevisualembedding AT wenzhengchen groupoptimizationformultiattributevisualembedding AT zhuohan groupoptimizationformultiattributevisualembedding AT mingyishi groupoptimizationformultiattributevisualembedding AT yanirkleiman groupoptimizationformultiattributevisualembedding AT danielcohenor groupoptimizationformultiattributevisualembedding AT baoquanchen groupoptimizationformultiattributevisualembedding AT yangyanli groupoptimizationformultiattributevisualembedding |
_version_ |
1725306198094249984 |