Mining Multimedia and Complex Data

Proceedings of the 21st International Conference on Pattern Recognition, 2012

This thesis presents a comparative study of different classification methodologies for the task of fine-art genre classification. The problem of painting classification involves classifying new unknown paintings among different art genres. Two-level comparative study is performed for this classification problem. The first level reviews the performance of discriminative vs. generative models while the second level touches the features aspect of the paintings and compares Semantic-level features vs low-level and intermediate-level features present in the painting. Three models are studied and compared, namely-1) A Discriminative model using a Bag-of-Words (BoW) approach; 2) A Generative model using BoW; 3) Discriminative model using Semantic-level features. Various experiments and techniques like Bag of Words model, Topic models and Classeme features are employed to get insights into potential of these automatic classification techniques for painting styles.

Journal of The American Society for Information Science and Technology, 2001

This article presents exploratory research evaluating a conceptual structure for the description of visual content of images. The structure, which was developed from empirical research in several fields (e.g., Computer Science, Psychology, Information Studies, etc.), classifies visual attributes into a "Pyramid" containing four syntactic levels (type/technique, global distribution, local structure, composition), and six semantic levels (generic, specific, and abstract levels of both object and scene, respectively). Various experiments are presented, which address the Pyramid's ability to achieve several tasks:

2008

Building product information is required during the architectural design and technical design. The common access to the technical information system is the multi-criteria search mode. This search mode is adapted to the situation where an architect has a precise demand of information. But most of the time, the architect looks for ideas and wants to obtain many illustrations of product uses. Therefore, the system has to propose another search mode adapted to the situation where the demand is still fuzzy. Considering that the architect has the capacity to think with image and that an image can generate easily ideas, then a search by images seems to be suitable to the situation where an architect looks for ideas. The web is an inexhaustible resource of images we can exploit to provision an image database on a specific area. The system we propose allows making building product information search with images extracted from the web. This article presents the method used to extract images from web sites of French building product companies and how these images are used in an interactive and progressive image retrieval process.

2010 10th International Conference on Intelligent Systems Design and Applications, 2010

Inspite of the huge amounts of image data on the web, mining image data from the web is paid less attention than mining text data, since treating the semantics of images is much more difficult. This paper introduces a new system to mine visual knowledge on the web that aims to build a Domain Oriented Image Directory by using the Earth Mover's Distance and Color signatures. Instead of using a flat classifier to combine text and image classification, the system suggests dividing the classification task into smaller classification problems corresponding to the branches in the classification hierarchy. Thus a multiple classifier system is presented. This paper illustrates the suggested system and discusses each of its components. Extensive experiments were conducted to test the system and also to compare it with commercial search engines. By the experiments we show that the proposed system accuracy outperforms the mostly used commercial search engines .

Advanced Engineering Informatics, 2018

The purpose of this research was to develop a case-based method for analyzing online customer reviews and extracting customer preferences through an integration of text mining and Kansei Engineering (KE) in an effort to achieve conceptual data-driven design automation and to successfully identify future trends in a particular consumer product. Design/Methodology/Approach This study's model merges text mining and KE to extract key descriptive Kansei terminology according to actual customer reviews and use it to forecast consumer preferred product design while reducing certain repetitious tasks of designers. This work first collects online product reviews using text mining. Then, through the application of KE, the customer-preferred design components are identified and incorporated into the product design specifications. Finally, an Application Programming Interface (API) is developed to automatically generate a CAD preliminary design. Case Study A road bike case study is provided to demonstrate the practicality of proposed method. The online reviews are collected from Amazon.com. The related design elements are classified into six key components which can be modified in the proposed conceptual design automation system. Originality/value This is the first paper that has applied text mining and KE for use in product development. This work can also reduce the time and cost of product design through the automation of repetitive design tasks. The conceptual design automation system is valuable for designers wishing to identify customer needs and to generate engineering drawings in a timely manner without significant repetition during the design process.

ACM International Conference on Multimedia Retrieval, ICMR 2012

Many objects have repetitive elements, and finding repetitive patterns facilitates object recognition and numerous applications. We devise a representation to describe configurations of repetitive elements. By modeling spatial configurations, visual patterns are more discriminative than local features, and are able to tackle with object scaling, rotation, and deformation. We transfer the pattern discovery problem into finding frequent subgraphs from a graph, and exploit a graph mining algorithm to solve this problem. Visual patterns are then exploited in architecture image classification and product image retrieval, based on the idea that visual pattern can describe elements conveying architecture styles and emblematic motifs of brands. Experimental results show that our pattern discovery approach has promising performance and is superior to the conventional bag-of-words approach.

1999

In this paper, we study computational models and techniques to combine textural and image features for classification of images on Internet. A framework is given to index images on the basis of textural, pictorial and composite information. The scheme makes use of weighted document terms and color invariant image features to obtain a high-dimensional similarity descriptor to be used as an index. Based on supervised learning, the k-nearest neighbor classifier is used to organize images into semantically meaningful groups of Internet images. Internet images are first classified into photographical and synthetical images. After classifying images into photographical and synthetical images, we further classify photographical images into portraits and non-portraits. Further, synthetical images are classified into button and non-button images.

NDDL, 2001

In this work, we present a system for categorizing photographs based on the text of their captions. The system has been developed as a part of the system CODI, an e-commerce application for an Spanish news agency. The categorization system makes able to the user the personalization of their information interests, improving search possibilities in the CODI application. Our approach for photograph categorization is based on linear text classifiers and Web mining programs, specially selected due to their suitability for industrial applications. The evaluation of our categorization system has shown that it meets the efficiency and effectiveness requirements of the e-commerce application.

A lot of research shown that designing an architectural solution requires introducing external references other than those existing in the design problem. In our case we choose photographic images to illustrate these references because an image illustrates the graphic part of an architectural concept. For this the information transmitted by photographic image is often easier to interpret by the designer. To assist the design process, we are working on the construction of a reference image database where the designer will find various solutions to his design problem. We have therefore proposed a method for indexing this image database. The following paper tackles the particular question of semantic meaning applied to architectural works. More precisely, we present the results obtained through an experiment following the Chi-square statistic rule, whose objective was to validate the weight associated with the architectural concepts illustrated according to their visual features in ima...

Journal of Mechanical Design, 2017

With the advent of the big-data era, massive information stored in electronic and digital forms on the internet become valuable resources for knowledge discovery in engineering design. Traditional document retrieval method based on document indexing focuses on retrieving individual documents related to the query, but is incapable of discovering the various associations between individual knowledge concepts. Ontology-based technologies, which can extract the inherent relationships between concepts by using advanced text mining tools, can be applied to improve design information retrieval in the large-scale unstructured textual data environment. However, few of the public available ontology database stands on a design and engineering perspective to establish the relations between knowledge concepts. This paper develops a “WordNet” focusing on design and engineering associations by integrating the text mining approaches to construct an unsupervised learning ontology network. Subsequent...