Bag-of-visual words (BoVW) is one of the most popular image representations in the image classification. Many features (also called visual words or codebooks) are generated to create the vector of images. However, similar to term in the text classification, these features may be relevant or irrelevant which affect the accuracy of classification. Many global weighting schemes (e.g. inverse document frequency) have been proposed to detect the relevant features. These global weighting schemes are based on document frequency (DF) in which the features will have the same weight when they have the same DF. This condition leads to reduce the discriminative power of features. Therefore, this study proposes a global weighting scheme based on intra-class and inter-class term distributions. The experiment was conducted by comparing the proposed method with the state-of-the-art global weighting schemes called inverse gravity moment (IGM) and modified inverse document frequency (mIDF). The evaluation of these weighting schemes is performed on the BoVW based image classification. Support vector machine (SVM) is used as a classifier to evaluate the methods on several benchmark datasets. By using the statistical analysis, such as Friedman nonparametric test, the proposed global weighting scheme outperforms the state-of-the-art global weighting schemes.
Rumpun Ilmu
Teknik Elektro
Bahasa Asli/Original Language
English
Level
Internasional
Status
Dokumen Karya
No
Judul
Tipe Dokumen
Aksi
1
2018 a global weighting scheme based on intra-class and inter-class term distributions in bag-of-visual words images classification-ilovepdf-compressed.pdf
[PAK] Full Dokumen
2
15 Similarity A global weighting scheme based on intra-class and inter-class term distributions in bag-of-visual words image classification.pdf
[PAK] Cek Similarity
3
18 Suket Publikasi PAK_A global weighting scheme.pdf
Dokumen Pendukung Karya Ilmiah (Hibah, Publikasi, Penelitian, Pengabdian)
