Researchers have developed a multimodal generative framework to aid the traditional Batak Ulos weaving industry by enabling controllable and culturally faithful motif generation. The system fine-tunes Stable Diffusion XL using LoRA and integrates with LLaMA 1.5-7B, employing four conditioning mechanisms: text, image, representation, and semantic map via ControlNet. An ablation study revealed that a combination of text, image, and semantic map conditioning yielded the best results for FID and CLIP Score, while text, image, and representation offered a balanced approach. User evaluations confirmed statistically significant positive acceptance, and a web-based prototype was created as a digital design tool. AI
IMPACT This research demonstrates a novel application of multimodal AI for cultural heritage preservation, potentially enabling new tools for artisans.
RANK_REASON The cluster describes a research paper detailing a novel application of a fine-tuned AI model for cultural heritage preservation. [lever_c_demoted from research: ic=1 ai=1.0]
- Batak Ulos
- Clip Score
- ControlNet
- Fréchet inception distance
- Humasak Simanjuntak
- Latent Diffusion Model
- LLaMA 1.5-7B
- LoRA
- Stable Diffusion XL
- Structural Similarity Index Measure
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →