Ancient Javanese Manuscript Reconstruction Using Generative Adversarial Network with StarGAN v2 Variations
DOI:
https://doi.org/10.34148/teknika.v14i1.1182Keywords:
Ancient Javanese Manuscripts, Image Reconstruction, Generative Adversial Network, StarGAN v2Abstract
Ancient Javanese manuscripts are part of Indonesia's cultural heritage; most of them are usually in bad condition due to the age and environmental surroundings. This paper presents a manuscript reconstruction using the Generative Adversarial Network model, using the variation of StarGAN v2. The primary objective of this research is to assist philologists in reconstructing damaged manuscripts more efficiently, reducing the time and effort compared to manual reconstruction methods. The training for 100 epochs is performed by the model in order to generate the reconstruction image closest to ground truth. This study is done on a dataset that consists of a set of damaged manuscript images. In this dataset, 80% is for training, 20% is for validation, and 10 images are used for testing. Quality assessment will be made on image outputs during training, based on PSNR, SSIM, and LPIPS metrics. The results indicate that the PSNR increases from 16.1234 dB at the 50th epoch to 17.5588 dB at the 100th epoch, while the SSIM increases from 0.8374 to 0.8519, showing a strong improvement in image quality. Despite the LPIPS having a very slight increase from 0.1020 to 0.1051, this evidences that the model can be further improved. Overall, this study demonstrates that the StarGAN v2 model is effective in reconstructing ancient Javanese manuscripts-a great contribution to the field of cultural heritage preservation using modern technology.
Downloads
References
[1] T. Asrianti and P. Y. Fauziah, “Pendampingan Belajar Aksara Jawa dalam Upaya Pelestarian Budaya Jawa,” Abdi J. Pengabdi. dan Pemberdaya. Masy., 2023, doi: 10.24036/abdi.v5i3.472.
[2] N. Nofrizal, “Pelestarian Manuskrip Kuno Melayu Nusantara Perspektif Industries,” Al-Adyan J. Stud. Lintas Agama, 2020, doi: 10.24042/ajsla.v15i2.6110.
[3] A. R. Himamunanto, “Restorasi Digital Pada Model Kerusakan Citra Aksara Jawa Cetak,” J. Teknol. Informasi-Aiti |, 2016.
[4] R. Raut, A. Devkar, P. S. Borkar, R. Deoghare, and S. Kolambe, “Generative Adversial Network Approach for Cartoonifying image using CartoonGAN,” J. Electr. Syst., 2023, doi: 10.52783/jes.666.
[5] Y. Choi, Y. Uh, J. Yoo, and J. W. Ha, “StarGAN v2: Diverse Image Synthesis for Multiple Domains,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2020. doi: 10.1109/CVPR42600.2020.00821.
[6] K. Holmes, P. Sharma, and S. Fernandes, “Facial skin disease prediction using StarGAN v2 and transfer learning,” Intell. Decis. Technol., 2023, doi: 10.3233/IDT-228046.
[7] Y. Assael et al., “Restoring and attributing ancient texts using deep neural networks,” Nature, 2022, doi: 10.1038/s41586-022-04448-z.
[8] A. Mumuni and F. Mumuni, “Data augmentation: A comprehensive survey of modern approaches,” Array. 2022. doi: 10.1016/j.array.2022.100258.
[9] C. Shorten and T. M. Khoshgoftaar, “A survey on Image Data Augmentation for Deep Learning,” J. Big Data, 2019, doi: 10.1186/s40537-019-0197-0.
[10] B. Han and M. Hu, “The Facial Expression Data Enhancement Method Induced by Improved StarGAN V2,” Symmetry (Basel)., 2023, doi: 10.3390/sym15040956.
[11] S. Park and Y. G. Shin, “Generative residual block for image generation,” Appl. Intell., 2022, doi: 10.1007/s10489-021-02858-6.
[12] L. Xinwei, G. Jinlin, D. Jinshen, and L. Songyang, “Generating Constrained Multi-target Scene Images Using Conditional SinGAN,” in 2021 IEEE 6th International Conference on Intelligent Computing and Signal Processing, ICSP 2021, 2021. doi: 10.1109/ICSP51882.2021.9408686.
[13] Y. Li, N. Xiao, and W. Ouyang, “Improved generative adversarial networks with reconstruction loss,” Neurocomputing, 2019, doi: 10.1016/j.neucom.2018.10.014.
[14] K. Janocha and W. M. Czarnecki, “On loss functions for deep neural networks in classification,” Schedae Informaticae, 2016, doi: 10.4467/20838476SI.16.004.6185.
[15] H. Zhao, O. Gallo, I. Frosio, and J. Kautz, “Loss Functions for Image Restoration With Neural Networks,” IEEE Trans. Comput. Imaging, 2017, doi: 10.1109/TCI.2016.2644865.
[16] National Instruments Australia, “Peak Signal-to-Noise ratio as an image quality metric,” Natl. Instruments, 2013.
[17] H. B. Sumarna, E. Utami, and A. D. Hartanto, “Tinjauan Literatur Sistematik tentang Structural Similarity Index Measure untuk Deteksi Anomali Gambar,” Creat. Inf. Technol. J., 2021, doi: 10.24076/citec.2020v7i2.248.
[18] H. Park and S. Park, “Improving Monocular Depth Estimation with Learned Perceptual Image Patch Similarity-Based Image Reconstruction and Left–Right Difference Image Constraints,” Electron., 2023, doi: 10.3390/electronics12173730.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Teknika
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
