Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Improving Web image search by bag-based reranking.

Lixin Duan¹, Wen Li, Ivor Wai-Hung Tsang

¹School of Computer Engineering, Nanyang Technological University, Singapore.

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society

|June 11, 2011

Summary

This study introduces a novel bag-based reranking framework for large-scale text-based image retrieval (TBIR). The proposed generalized multi-instance (GMI) learning approach, GMI-SVM, enhances image retrieval accuracy by effectively labeling image instances within bags.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

MUFOLD-DB: a processed protein structure database for protein structure prediction and analysis.

BMC genomics·2015

Same author

Advances in translational bioinformatics facilitate revealing the landscape of complex disease mechanisms.

BMC bioinformatics·2015

Same author

The I-TASSER Suite: protein structure and function prediction.

Nature methods·2014

Same author

Genome-wide expression analysis of soybean NF-Y genes reveals potential function in development and drought response.

Molecular genetics and genomics : MGG·2014

Same author

Classification of lung cancer using ensemble-based feature selection and machine learning methods.

Molecular bioSystems·2014

Same author

Resveratrol possesses protective effects in a pristane-induced lupus mouse model.

PloS one·2014

Area of Science:

Computer Science
Information Retrieval
Machine Learning

Background:

Traditional text-based image retrieval (TBIR) often requires reranking using visual features post-initial search.
Existing methods face challenges in effectively utilizing both textual and visual information for reranking large datasets.

Purpose of the Study:

To propose a new bag-based reranking framework for large-scale text-based image retrieval.
To develop a generalized multi-instance (GMI) learning approach, GMI-SVM, for improved image reranking.
To introduce an automatic bag annotation method for training the GMI learning model.

Main Methods:

Clustering relevant images using textual and visual features to form 'bags' of 'instances'.
Formulating the problem as a generalized multi-instance (GMI) learning task.

Related Experiment Videos

Developing GMI-SVM to propagate labels from bag to instance level for enhanced retrieval.

Implementing a bag ranking method for automatic annotation of training bags.

Main Results:

The proposed bag-based framework with automatic annotation achieves superior performance on the NUS-WIDE dataset compared to existing reranking methods.
GMI-SVM demonstrates improved retrieval accuracy, especially when trained with manually labeled bags from relevance feedback.
The GMI learning setting effectively addresses label ambiguities in positive and negative bags.

Conclusions:

The proposed bag-based reranking framework offers a significant advancement for large-scale text-based image retrieval.
GMI-SVM provides a robust solution for instance-level labeling in a GMI setting, boosting retrieval performance.
Automatic bag annotation is a viable strategy for training, with manual labeling offering further performance gains.