Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data

Giorgio Vassallo, Marco La Cascia, Luigi Gallo, Giovanni Pilato, Filippo Vella

Risultato della ricerca: Conference contribution

Abstract

The main drawback of a detailed representation of visual content, whatever is its origin, is that significant features are very high dimensional. To keep the problem tractable while preserving the semantic content, a dimensionality reduction of the data is needed. We propose the Random Projection techniques to reduce the dimensionality. Even though this technique is sub-optimal with respect to Singular Value Decomposition its much lower computational cost make it more suitable for this problem and in particular when computational resources are limited such as in mobile terminals. In this paper we present the use of a 'conceptual' space, automatically induced from data, to perform automatic image annotation. Images are represented by visual features based on color and texture and arranged as histograms of visual terms and bigrams to partially preserve the spatial information [1]. Using a set of annotated images as training data, the matrix of visual features is built and dimensionality reduction is performed using the Random Projection algorithm. A new unannotated image is then projected into the dimensionally reduced space and the labels of the closest training images are assigned to the unannotated image itself. Experiments on large real collection of images showed that the approach, despite of its low computational cost, is very effective.
Lingua originaleEnglish
Titolo della pubblicazione ospiteProceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018
Pagine464-471
Numero di pagine8
Stato di pubblicazionePublished - 2018

Fingerprint

Singular value decomposition
Labels
Costs
Textures
Semantics
Color
Experiments

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Computer Vision and Pattern Recognition

Cita questo

Vassallo, G., La Cascia, M., Gallo, L., Pilato, G., & Vella, F. (2018). Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data. In Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018 (pagg. 464-471)

Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data. / Vassallo, Giorgio; La Cascia, Marco; Gallo, Luigi; Pilato, Giovanni; Vella, Filippo.

Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. 2018. pag. 464-471.

Risultato della ricerca: Conference contribution

Vassallo, G, La Cascia, M, Gallo, L, Pilato, G & Vella, F 2018, Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data. in Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. pagg. 464-471.
Vassallo G, La Cascia M, Gallo L, Pilato G, Vella F. Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data. In Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. 2018. pag. 464-471
Vassallo, Giorgio ; La Cascia, Marco ; Gallo, Luigi ; Pilato, Giovanni ; Vella, Filippo. / Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data. Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018. 2018. pagg. 464-471
@inproceedings{85c7b1fd8c214b2abc6517671a364cc2,
title = "Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data",
abstract = "The main drawback of a detailed representation of visual content, whatever is its origin, is that significant features are very high dimensional. To keep the problem tractable while preserving the semantic content, a dimensionality reduction of the data is needed. We propose the Random Projection techniques to reduce the dimensionality. Even though this technique is sub-optimal with respect to Singular Value Decomposition its much lower computational cost make it more suitable for this problem and in particular when computational resources are limited such as in mobile terminals. In this paper we present the use of a 'conceptual' space, automatically induced from data, to perform automatic image annotation. Images are represented by visual features based on color and texture and arranged as histograms of visual terms and bigrams to partially preserve the spatial information [1]. Using a set of annotated images as training data, the matrix of visual features is built and dimensionality reduction is performed using the Random Projection algorithm. A new unannotated image is then projected into the dimensionally reduced space and the labels of the closest training images are assigned to the unannotated image itself. Experiments on large real collection of images showed that the approach, despite of its low computational cost, is very effective.",
keywords = "Conceptual-spaces, Image-semantics, Random-projection",
author = "Giorgio Vassallo and {La Cascia}, Marco and Luigi Gallo and Giovanni Pilato and Filippo Vella",
year = "2018",
language = "English",
isbn = "978-1-5386-9385-8",
pages = "464--471",
booktitle = "Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018",

}

TY - GEN

T1 - Automatic Image Annotation Using Random Projection in a Conceptual Space Induced from Data

AU - Vassallo, Giorgio

AU - La Cascia, Marco

AU - Gallo, Luigi

AU - Pilato, Giovanni

AU - Vella, Filippo

PY - 2018

Y1 - 2018

N2 - The main drawback of a detailed representation of visual content, whatever is its origin, is that significant features are very high dimensional. To keep the problem tractable while preserving the semantic content, a dimensionality reduction of the data is needed. We propose the Random Projection techniques to reduce the dimensionality. Even though this technique is sub-optimal with respect to Singular Value Decomposition its much lower computational cost make it more suitable for this problem and in particular when computational resources are limited such as in mobile terminals. In this paper we present the use of a 'conceptual' space, automatically induced from data, to perform automatic image annotation. Images are represented by visual features based on color and texture and arranged as histograms of visual terms and bigrams to partially preserve the spatial information [1]. Using a set of annotated images as training data, the matrix of visual features is built and dimensionality reduction is performed using the Random Projection algorithm. A new unannotated image is then projected into the dimensionally reduced space and the labels of the closest training images are assigned to the unannotated image itself. Experiments on large real collection of images showed that the approach, despite of its low computational cost, is very effective.

AB - The main drawback of a detailed representation of visual content, whatever is its origin, is that significant features are very high dimensional. To keep the problem tractable while preserving the semantic content, a dimensionality reduction of the data is needed. We propose the Random Projection techniques to reduce the dimensionality. Even though this technique is sub-optimal with respect to Singular Value Decomposition its much lower computational cost make it more suitable for this problem and in particular when computational resources are limited such as in mobile terminals. In this paper we present the use of a 'conceptual' space, automatically induced from data, to perform automatic image annotation. Images are represented by visual features based on color and texture and arranged as histograms of visual terms and bigrams to partially preserve the spatial information [1]. Using a set of annotated images as training data, the matrix of visual features is built and dimensionality reduction is performed using the Random Projection algorithm. A new unannotated image is then projected into the dimensionally reduced space and the labels of the closest training images are assigned to the unannotated image itself. Experiments on large real collection of images showed that the approach, despite of its low computational cost, is very effective.

KW - Conceptual-spaces

KW - Image-semantics

KW - Random-projection

UR - http://hdl.handle.net/10447/389736

UR - http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8693683

M3 - Conference contribution

SN - 978-1-5386-9385-8

SP - 464

EP - 471

BT - Proceedings - 14th International Conference on Signal Image Technology and Internet Based Systems, SITIS 2018

ER -