Image tag refinement towards low-rank, content-tag prior and error sparsity

Zhu, Guangyu; Yan, Shuicheng; Ma, Yi

doi:10.1145/1873951.1874028

Cited by 225 publications

(168 citation statements)

References 20 publications

Supporting

Mentioning

168

Contrasting

Order By: Relevance

“…this method also account for the calculated stabletag count through its occurrence.  Zhu et al [16] proposed RobustPCA method [6]. This method is based on analyzing the main powerful factors, D matrix (tag and image) is factorization by analysis of low Ran k decomposition with error scarcity and it is D=Ď+E in which Ď has a low rank constraint based on the nuclear norm, and E is an error mat rix with l 1 -normsparsity constraint.…”

Section:  mentioning

confidence: 99%

A Review of Methods of Instance-based Automatic Image Annotation

Derakhshan¹,

Maihami²

2016

IJISA

View full text Add to dashboard Cite

Abstract-Today, to use automatic image annotation in order to fill the semantic gap between low level features of images and understanding their information in retriev ing process has become popular. Since automatic image annotation is crucial in understanding digital images several methods have been proposed to automatically annotate an image. One of the most important of these methods is instance-based image annotation. As these methods are vastly used in this paper, the most impo rtant instance-based image annotation methods are analy zed. First of all the main parts of instance-based automatic image annotation are analyzed. Afterwards, the main methods of instance-based automatic image annotation are reviewed and co mpared based on various features. In the end the most important challenges and open-ended fields in instance-based image annotation are analyzed.

show abstract

Section:  mentioning

confidence: 99%

A Review of Methods of Instance-based Automatic Image Annotation

Derakhshan¹,

Maihami²

2016

IJISA

View full text Add to dashboard Cite

show abstract

“…Recently, a promising work for tag refinement was presented in [24], where Zhu et al proposed to decompose a user-provided tag matrix into a low-rank refined matrix and a sparse error matrix. The work in [24] is effective for a global tag refinement; however, it may not be appropriate to our problem here as each entry in our tag matrices represents a score and our objective is to tune the scores locally (a global tag refinement may obscure the experimental results of our proposed approaches).…”

Section: Tag Refinementmentioning

confidence: 99%

“…The work in [24] is effective for a global tag refinement; however, it may not be appropriate to our problem here as each entry in our tag matrices represents a score and our objective is to tune the scores locally (a global tag refinement may obscure the experimental results of our proposed approaches). As a result, we modified the approach of learning class-specific weights for the searching of visually similar images in [25] to fit our needs.…”

Section: Tag Refinementmentioning

confidence: 99%

Graph-based semi-supervised learning with multi-modality propagation for large-scale image datasets

Lee

Hsieh

et al. 2013

Journal of Visual Communication and Image Representation

View full text Add to dashboard Cite

Semi-supervised learning (SSL) is widely-used to explore the vast amount of unlabeled data in the world. Over the decade, graph-based SSL becomes popular in automatic image annotation due to its power of learning globally based on local similarity. However, recent studies have shown that the emergence of large-scale datasets challenges the traditional methods. On the other hand, most previous works have concentrated on single-label annotation, which may not describe image contents well. To remedy the deficiencies, this paper proposes a new graph-based SSL technique with multi-label propagation, leveraging the distributed computing power of the MapReduce programming model. For high learning performance, the paper further presents both a multi-layer learning structure and a tag refinement approach, where the former unifies both visual and textual information of image data during learning, while the latter simultaneously suppresses noisy tags and emphasizes the other tags after learning. Experimental results based on a medium-scale and a large-scale image datasets show the effectiveness of the proposed methods.

show abstract

“…Wang et al [28] address the tag optimization problem given the tagging labels of images obtained by a classifier; Zhu et al [31] propose a classification method to address the noisy tagging problem; neither of the methods explicitly tackles the incomplete tagging problem. Liu et al [18] propose a solution that denoises the noisy tagging and enriches the incomplete tagging, while their enrichment can only add synonyms and hypernyms, and needs the availability and assistance from WordNet.…”

Section: Related Workmentioning

confidence: 99%

“…Liu et al [18] propose a solution that denoises the noisy tagging and enriches the incomplete tagging, while their enrichment can only add synonyms and hypernyms, and needs the availability and assistance from WordNet. In addition, the classification solutions in [18] and [31] are not inductive in learning.…”

Section: Related Workmentioning

confidence: 99%

Multi-view learning from imperfect tagging

Zhang

2012

Proceedings of the 20th ACM International Conference on Multimedia

View full text Add to dashboard Cite

In many real-world applications, tagging is imperfect: incomplete, inconsistent, and error-prone. Solutions to this problem will generate societal and technical impacts. In this paper, we investigate this arguably new problem: learning from imperfect tagging. We propose a general and effective learning scheme called the Multiview Imperfect Tagging Learning (MITL) to this problem. The main idea of MITL lies in extracting the information of the imperfectly tagged training dataset from multiple views to differentiate the data points in the role of classification. Further, a novel discriminative classification method is proposed under the framework of MITL, which explicitly makes use of the given multiple labels simultaneously as an additional feature to deliver a more effective classification performance than the existing literature where one label is considered at a time as the classification target while the rest of the given labels are completely ignored at the same time. The proposed methods can not only complete the incomplete tagging but also denoise the noisy tagging through an inductive learning. We apply the general solution to the problem with a more specific context -imperfect image annotation, and evaluate the proposed methods on a standard dataset from the related literature. Experiments show that they are superior to the peer methods on solving the problem of learning from imperfect tagging in cross-media.

show abstract

Image tag refinement towards low-rank, content-tag prior and error sparsity

Cited by 225 publications

References 20 publications

A Review of Methods of Instance-based Automatic Image Annotation

A Review of Methods of Instance-based Automatic Image Annotation

Graph-based semi-supervised learning with multi-modality propagation for large-scale image datasets

Multi-view learning from imperfect tagging

Contact Info

Product

Resources

About