A feature-word-topic model for image annotation

Nguyen, Cam-Tu; Kaothanthong, Natsuda; Phan, Xuan-Hieu; Tokuyama, Takeshi

doi:10.1145/1871437.1871652

Cited by 9 publications

(6 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…But they are clearly different because our approach imposes the regularization over the topic distribution instead of the latent variables and moreover our approach deals multiple modalities and makes use of additional visual similarity to formulate the regularization term. Our approach is also different from topic models for image annotation [4,2,28,29]: The image tagging problem in our paper is more challenging than image annotation as aforementioned. and moreover the proposed regularized LDA aims to impose the consistency of tags within similar images while they are supervised algorithms [28,29] or aim to find common topics shared by tags and visual contents.…”

Section: Related Workmentioning

confidence: 99%

“…The automatic image tagging or annotation problem is usually regarded as an image classification task. Typical techniques [2,4,7,9,11,12,14,16,20,28,29,33] usually learn a generative/discriminative multi-class classifier from the training data, to construct a mapping function from low level features extracted from the images to tags, and then predict the annotations or tags for the test images. Later, a more precise formulation is presented to regard it as a multi-label classification problem by exploring the relations between multiple labels [15,30].…”

Section: Related Workmentioning

confidence: 99%

“…Our approach is also different from topic models for image annotation [4,2,28,29]: The image tagging problem in our paper is more challenging than image annotation as aforementioned. and moreover the proposed regularized LDA aims to impose the consistency of tags within similar images while they are supervised algorithms [28,29] or aim to find common topics shared by tags and visual contents. This paper is different from the short version [37] because we present a formal derivation of our approach, introduce a new inference approach conduct more experiments, and particularly, we use the deep network structure to analyze the benefit of regularized LDA.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Image tag refinement by regularized latent Dirichlet allocation

Wang

Zhou

et al. 2014

Computer Vision and Image Understanding

View full text Add to dashboard Cite

Tagging is nowadays the most prevalent and practical way to make images searchable. However, in reality many manually-assigned tags are irrelevant to image content and hence are not reliable for applications. A lot of recent efforts have been conducted to refine image tags. In this paper, we propose to do tag refinement from the angle of topic modeling and present a novel graphical model, regularized Latent Dirichlet Allocation (rLDA). In the proposed approach, tag similarity and tag relevance are jointly estimated in an iterative manner, so that they can benefit from each other, and the multi-wise relationships among tags are explored. Moreover, both the statistics of tags and visual affinities of images in the corpus are explored to help topic modeling. We also analyze the superiority of our approach from the deep structure perspective. The experiments on tag ranking and image retrieval demonstrate the advantages of the proposed method.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Image tag refinement by regularized latent Dirichlet allocation

Wang

Zhou

et al. 2014

Computer Vision and Image Understanding

View full text Add to dashboard Cite

show abstract

“…Akbas et al (Akbas and Vural, 2007) fused binary classifiers by learning a new meta classifier from categorymembered vectors, which are generated from the binary classifiers. Nguyen et al (Nguyen et al, 2010) proposed a feature-word-topic model in which one individual classifier is learned for each label based on visual features. By modeling topics of words, the authors then refine the results from binary classifiers to obtain topic-oriented annotation for later image retireval.…”

Section: Classification-based Approachmentioning

confidence: 99%

Cascade of Multi-Level Multi-Instance Classifiers for Image Annotation

Nguyen

Tokuyama

2011

Proceedings of the International Conference on Knowledge Discovery and Information Retrieval

View full text Add to dashboard Cite

Abstract:This paper introduces a new scheme for automatic image annotation based on cascading multi-level multiinstance classifiers (CMLMI). The proposed scheme employs a hierarchy for visual feature extraction, in which the feature set includes features extracted from the whole image at the coarsest level and from the overlapping sub-regions at finer levels. Multi-instance learning (MIL) is used to learn the "weak classifiers" for these levels in a cascade manner. The underlying idea is that the coarse levels are suitable for background labels such as "forest" and "city", while finer levels bring useful information about foreground objects like "tiger" and "car". The cascade manner allows this scheme to incorporate "important" negative samples during the learning process, hence reducing the "weakly labeling" problem by excluding ambiguous background labels associated with the negative samples. Experiments show that the CMLMI achieve significant improvements over baseline methods as well as existing MIL-based methods.

show abstract

“…By considering an image as an exam-ple and its instances as feature vectors extracted from subregions of the image, the problem of image annotation naturally fits the multi-instance setting. There were several studies (Carneiro et al, 2007;Nguyen et al, 2010;Yang et al, 2006) that have successfully applied MIL to the problem of image annotation. One disadvantage of those methods is that they treated instances (sub-regions in images) equally for every label.…”

Section: Introductionmentioning

confidence: 99%

A Multiple Instance Learning Approach to Image Annotation with Saliency Map

Nhung¹,

Nguyen

Jin

et al. 2013

Proceedings of the International Conference on Knowledge Discovery and Information Retrieval and the International Conference O

View full text Add to dashboard Cite

This paper presents a novel approach to image annotation based on multi-instance learning (MIL) and saliency map. Image Annotation is an automatic process of assigning labels to images so as to obtain semantic retrieval of images. This problem is often ambiguous as a label is given to the whole image while it may only corresponds to a small region in the image. As a result, MIL methods are suitable solutions to resolve the ambiguities during learning. On the other hand, saliency detection aims at detecting foreground/background regions in images. Once we obtain this information, labels and image regions can be aligned better, i.e., foreground labels (background labels) are more sensitive to foreground areas (background areas). Our proposed method, which is based on an ensemble of MIL classifiers from two views (background/foreground), improves annotation performance in comparison to baseline methods that do not exploit saliency information.

show abstract

A feature-word-topic model for image annotation

Cited by 9 publications

References 10 publications

Image tag refinement by regularized latent Dirichlet allocation

Image tag refinement by regularized latent Dirichlet allocation

Cascade of Multi-Level Multi-Instance Classifiers for Image Annotation

A Multiple Instance Learning Approach to Image Annotation with Saliency Map

Contact Info

Product

Resources

About