The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
A representação Bag-of-Visual-Words tornou-se recentemente popular para classificação de cenas. Porém, aprender as palavras visuais de maneira não supervisionada sofre com o problema quando se depara com essas manchas com aparências semelhantes correspondentes a conceitos semânticos distintos. Este artigo propõe uma nova estrutura de aprendizagem supervisionada, que visa aproveitar ao máximo as informações do rótulo para resolver o problema. Especificamente, a Modelagem de Mistura Gaussiana (GMM) é aplicada primeiramente para obter "interpretação semântica" de patches usando rótulos de cena. Cada cena induz uma densidade de probabilidade no espaço de características visuais de baixo nível, e os patches são representados como vetores de probabilidades de conceitos semânticos da cena posterior. E então o algoritmo Information Bottleneck (IB) é introduzido para agrupar os patches em "palavras visuais" de forma supervisionada, do ponto de vista das interpretações semânticas. Tal operação pode maximizar a informação semântica das palavras visuais. Uma vez obtidas as palavras visuais, a frequência de aparecimento das palavras visuais correspondentes em uma determinada imagem forma um histograma, que pode ser posteriormente utilizado na tarefa de categorização de cena através do classificador Support Vector Machine (SVM). Experimentos em um conjunto de dados desafiador mostram que as palavras visuais propostas executam melhor a tarefa de classificação de cena do que a maioria dos métodos existentes.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Shuoyan LIU, De XU, Songhe FENG, "Discriminating Semantic Visual Words for Scene Classification" in IEICE TRANSACTIONS on Information,
vol. E93-D, no. 6, pp. 1580-1588, June 2010, doi: 10.1587/transinf.E93.D.1580.
Abstract: Bag-of-Visual-Words representation has recently become popular for scene classification. However, learning the visual words in an unsupervised manner suffers from the problem when faced these patches with similar appearances corresponding to distinct semantic concepts. This paper proposes a novel supervised learning framework, which aims at taking full advantage of label information to address the problem. Specifically, the Gaussian Mixture Modeling (GMM) is firstly applied to obtain "semantic interpretation" of patches using scene labels. Each scene induces a probability density on the low-level visual features space, and patches are represented as vectors of posterior scene semantic concepts probabilities. And then the Information Bottleneck (IB) algorithm is introduce to cluster the patches into "visual words" via a supervised manner, from the perspective of semantic interpretations. Such operation can maximize the semantic information of the visual words. Once obtained the visual words, the appearing frequency of the corresponding visual words in a given image forms a histogram, which can be subsequently used in the scene categorization task via the Support Vector Machine (SVM) classifier. Experiments on a challenging dataset show that the proposed visual words better perform scene classification task than most existing methods.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E93.D.1580/_p
Copiar
@ARTICLE{e93-d_6_1580,
author={Shuoyan LIU, De XU, Songhe FENG, },
journal={IEICE TRANSACTIONS on Information},
title={Discriminating Semantic Visual Words for Scene Classification},
year={2010},
volume={E93-D},
number={6},
pages={1580-1588},
abstract={Bag-of-Visual-Words representation has recently become popular for scene classification. However, learning the visual words in an unsupervised manner suffers from the problem when faced these patches with similar appearances corresponding to distinct semantic concepts. This paper proposes a novel supervised learning framework, which aims at taking full advantage of label information to address the problem. Specifically, the Gaussian Mixture Modeling (GMM) is firstly applied to obtain "semantic interpretation" of patches using scene labels. Each scene induces a probability density on the low-level visual features space, and patches are represented as vectors of posterior scene semantic concepts probabilities. And then the Information Bottleneck (IB) algorithm is introduce to cluster the patches into "visual words" via a supervised manner, from the perspective of semantic interpretations. Such operation can maximize the semantic information of the visual words. Once obtained the visual words, the appearing frequency of the corresponding visual words in a given image forms a histogram, which can be subsequently used in the scene categorization task via the Support Vector Machine (SVM) classifier. Experiments on a challenging dataset show that the proposed visual words better perform scene classification task than most existing methods.},
keywords={},
doi={10.1587/transinf.E93.D.1580},
ISSN={1745-1361},
month={June},}
Copiar
TY - JOUR
TI - Discriminating Semantic Visual Words for Scene Classification
T2 - IEICE TRANSACTIONS on Information
SP - 1580
EP - 1588
AU - Shuoyan LIU
AU - De XU
AU - Songhe FENG
PY - 2010
DO - 10.1587/transinf.E93.D.1580
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E93-D
IS - 6
JA - IEICE TRANSACTIONS on Information
Y1 - June 2010
AB - Bag-of-Visual-Words representation has recently become popular for scene classification. However, learning the visual words in an unsupervised manner suffers from the problem when faced these patches with similar appearances corresponding to distinct semantic concepts. This paper proposes a novel supervised learning framework, which aims at taking full advantage of label information to address the problem. Specifically, the Gaussian Mixture Modeling (GMM) is firstly applied to obtain "semantic interpretation" of patches using scene labels. Each scene induces a probability density on the low-level visual features space, and patches are represented as vectors of posterior scene semantic concepts probabilities. And then the Information Bottleneck (IB) algorithm is introduce to cluster the patches into "visual words" via a supervised manner, from the perspective of semantic interpretations. Such operation can maximize the semantic information of the visual words. Once obtained the visual words, the appearing frequency of the corresponding visual words in a given image forms a histogram, which can be subsequently used in the scene categorization task via the Support Vector Machine (SVM) classifier. Experiments on a challenging dataset show that the proposed visual words better perform scene classification task than most existing methods.
ER -