Non-linear Multiclass SVM Classification Optimization using Large Datasets of Geometric Motif Image

Abstract

Support Vector Machine (SVM) with Radial Basis Functions (RBF) kernel is one of the methods frequently applied to nonlinear multiclass image classification. To overcome some constraints in the form of a large number of image datasets divided into nonlinear multiclass, there three stages of SVM-RBF classification process carried out i.e. 1) Determining the algorithms of feature extraction and feature value dimensions used, 2) Determining the appropriate kernel and parameter values, and 3) Using correct multiclass method for the training and testing processes. The OaO, OaA, and DAGSVM multi-class methods were tested on a large dataset of batik motif images whose geometric motifs with a variety of patterns and colors in each class and containing similar patterns in the motifs between the classes. DAGSVM has the advantage in classification accuracy value, i.e. 91%, but it takes longer during the training and testing processes.

Review on integration of ontology and deep learning in cultural heritage image retrieval

“…Research in the field of image retrieval is still being conducted today, due to the influence of the interconnected domains of the web, big data, ontology, and deep learning. What is done in Image Retrieval is the calculation of similarity in images, in content base image retrieval (CBIR) the search for similar images based on colour characteristics, image edges, texture, or based on spatial information [1]- [3]. The application of CBIR on web documents is increasingly inadequate because it has problems with big data.…”

“…Big data documents occur due to information that is continuously uploaded to the internet, so that documents have a very large volume of data with many classes, heterogeneous, and very fast growth in number, and have unstructured and semi-structured data forms [4]- [8]. Previous CBIR research with an image dataset in the cultural heritage domain had a high accuracy of 96.7% when applied to classification with two classes [9], in subsequent research with the same data and methods applied to multi-class classification decreased accuracy [3], [10], the same multi-class classification method tested with large data decreased accuracy and training time became very long [3]. In handling very large data waves and many classes, CBIR has developed from  ISSN: 2502-4752 Indonesian J Elec Eng & Comp Sci, Vol.…”

“…Ontologies can help identify relevant feature patterns in images, and can improve semantic understanding of image content [25]. For example, image retrieval research in recognising cultural heritage image domain, in this case Indonesian batik image [3], [9], [10]. Batik is an artwork of drawing motifs on fabric media, which is coloured by a waxresist dyeing process, and uses tools called "canting" and "cap" [26].…”

“…Batik is an artwork of drawing motifs on fabric media, which is coloured by a waxresist dyeing process, and uses tools called "canting" and "cap" [26]. CBIR research on classification to recognise Batik motifs with many datasets and classes has problems with the training speed and classification process [3], so in this case the concept of ontology is needed. Ontology can create a concept hierarchy by identifying role relationships between concepts related to batik image characteristics such as types of batik motifs, commonly used colours, fabric types, regions of origin, and other characteristics contained in batik images, so that users can search based on the categories specified in the ontology [27]- [29].…”

“…Research on the relationship between ontology and deep learning in recognising images as a cultural heritage domain is still developing, and analysis of various approaches and techniques is still needed. CBIR research on big data batik images with geometric motifs [3], [10], is currently still limited to deep learning research in the feature extraction or classification part to produce maximum image recognition accuracy. In general, an important problem in image retrieval is how to build a comprehensive, relevant, and precise ontology design with a deep understanding of the domain under study to support image retrieval.…”