Clustering OKU Timur Script Images using VGG Feature extraction and K-Means
This study focuses on the utilization of clustering models to group manuscript images from the OKU Timur region based on specific characteristics. OKU Timur is rich in cultural heritage, including a unique writing system known as the OKU Timur script. The development of intelligent systems technolog...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
LPPM ISB Atma Luhur
2024-12-01
|
| Series: | Jurnal Sisfokom |
| Subjects: | |
| Online Access: | https://jurnal.atmaluhur.ac.id/index.php/sisfokom/article/view/2292 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study focuses on the utilization of clustering models to group manuscript images from the OKU Timur region based on specific characteristics. OKU Timur is rich in cultural heritage, including a unique writing system known as the OKU Timur script. The development of intelligent systems technology can be employed to recognize the OKU Timur script. For this purpose, a dataset of OKU Timur script is needed, which will later be used for classifying script images. One of the challenges in preparing the dataset is grouping a large number of script image samples according to the number of characters. A proposed solution in this research is to automatically group script images by applying the K-Means algorithm. The dataset comprises 2,280 images, representing 19 characters and 228 variations with different diacritics. Features are extracted using the VGG16 model, which are then clustered with the K-Means algorithm. Clustering performance is evaluated based on the percentage of correctly grouped characters. For 19 groups (character count), the model achieves an accuracy of 82.6%. For 228 groups (variations and diacritics), it correctly groups 48.16% of characters. Despite the challenges, the results demonstrate the model’s potential for further refinement. This study’s contribution lies in introducing an efficient clustering approach for cultural manuscripts, supporting digital preservation, and advancing automatic recognition of the OKU Timur script. These efforts aim to preserve the script for future generations. |
|---|---|
| ISSN: | 2301-7988 2581-0588 |