Penerapan Low-Rank Adaptation (LoRA) pada Stable Diffusion untuk Generasi Citra Motif Batik Indonesia
DOI:
https://doi.org/10.24114/cess.v11i2.75575Keywords:
Low-Rank Adaptation (LoRA);, Stable Diffusion, Generasi Citra, Batik Indonesia;, Kecerdasan BuatanAbstract
The development of diffusion-based text-to-image models has created new opportunities for generating Indonesian batik motifs using artificial intelligence. This study aims to apply Low-Rank Adaptation (LoRA) to Stable Diffusion 1.5 to generate images of six Indonesian batik motifs: Parang, Ceplok, Kawung, Buketan, Lasem, and Mega Mendung. The dataset consists of 300 images that underwent a preprocessing stage before being trained using Kohya SS. The model was evaluated using the Frechet Inception Distance (FID) and the Contrastive Language–Image Pre-training (CLIP) Score. The results show that the Buketan motif achieved the best performance, with the lowest FID value of 24.81 and the highest CLIP Score of 0.351, while the Parang motif obtained the highest FID value of 41.52 and the lowest CLIP Score of 0.286. These findings indicate that the LoRA method is capable of generating Indonesian batik motif images with good visual quality and semantic alignment across several motif categories, although its performance still varies depending on the characteristics of each motif.
Downloads
References
[1] A. E. Minarno, I. Soesanti, and H. A. Nugroho, “Dataset of Batik Nitik Sarimbit 120,” Data Br., p. 110671, 2024, doi: 10.1016/j.dib.2024.110671.
[2] V. J. Basiroen, “Batik Lasem motifs dataset: Gunung Ringgit, Kricak, Latohan, Nyuk Pitu, Seritan,” Data Br., 2025, doi: 10.1016/j.dib.2025.110671.
[3] R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-resolution image synthesis with latent diffusion models,” in CVPR 2022, 2022, pp. 10684–10695. [Online]. Available: https://arxiv.org/abs/2112.10752
[4] E. J. Hu et al., “LoRA: Low-rank adaptation of large language models,” in ICLR 2022, 2022. [Online]. Available: https://arxiv.org/abs/2106.09685
[5] Y. Mao, Y. Ge, Y. Fan, and Y. Gao, “A Survey on LoRA of Large Language Models,” Front. Comput. Sci., vol. 19, no. 7, 2024, doi: 10.1007/s11704-024-40663-9.
[6] M. T. Yılmaz, A. K. Erdogmus, D. Bayar, and G. Çökünlü, “Synthetic Image Generation with Stable Diffusion and Trained LoRA Model for Industrial Areas,” in 2024 International Conference on Artificial Intelligence and Signal Processing (ASYU), 2024. doi: 10.1109/ASYU62119.2024.10757052.
[7] M. F. Sutedy and N. N. Qomariyah, “Text to Image Latent Diffusion Model with DreamBooth Fine Tuning for Automobile Image Generation,” in 2022 5th International Seminar on Research of Information Technology and Intelligent Systems (ISRITI), Yogyakarta, Indonesia, 2022, pp. 343–348. doi: 10.1109/ISRITI56927.2022.10052908.
[8] J. Liu and Y. Huang, “LoRA-Fine-Tuned Latent Diffusion for High-Fidelity Digitization of Classic Mongolian Patterns,” Appl. Sci., vol. 16, no. 1, p. 11, 2026, doi: 10.3390/app16010011.
[9] A. Lv and J. Yen, “A Comparative Analysis of Different Versions of Stable Diffusion,” in 2025 IEEE International Conference on Advanced Visual and Signal-Based Systems (AVSS), Tainan, Taiwan, 2025. doi: 10.1109/AVSS65446.2025.11149921.
[10] I. Adegoke, D. Izenyi, and L. Dilli, “Efficiently Fine-tuning Large Language Model: LoRA Approach,” 2024, Zenodo. doi: 10.5281/zenodo.11312792.
[11] K. Schall, K. U. Barthel, N. Hezel, and A. Moelle, “A comprehensive approach to improving CLIP-based image retrieval while maintaining joint-embedding alignment,” Inf. Syst., vol. 134, p. 102581, 2025, doi: 10.1016/j.is.2025.102581.
[12] N. Chen, S. Villar, and S. Hayou, “Learning Rate Scaling across LoRA Ranks and Transfer to Full Finetuning,” in Proceedings of the 43rd International Conference on Machine Learning (ICML), 2026.
[13] J. V. M. Marques, W. S. Santos, A. I. M. Luz, R. L. Araújo, A. D. de Sousa, and J. L. R. Sarmento, “U-Net + Transformer Encoder and LoRA Adaptation for Rural Sheep Segmentation,” in 2025 38th SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), Salvador, Brazil, 2025. doi: 10.1109/SIBGRAPI67909.2025.11223404.
[14] D. A. Chan and S. P. Sithungu, “Evaluating the Suitability of Inception Score and Fréchet Inception Distance as Metrics for Quality and Diversity in Image Generation,” in Proceedings of the 2024 7th International Conference on Computational Intelligence and Intelligent Systems (CIIS ’24), 2025, pp. 79–85. doi: 10.1145/3708778.3708790.
[15] J. Hessel, A. Holtzman, M. Forbes, R. Le Bras, and Y. Choi, “CLIPScore: A Reference-free Evaluation Metric for Image Captioning,” in Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2021, pp. 7514–7528. doi: 10.18653/v1/2021.emnlp-main.595.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 CESS (Journal of Computer Engineering, System and Science)
This work is licensed under a Creative Commons Attribution 4.0 International License.
