Semantic Seeker
Semantic Seeker was showcased at Taiwan Design Week. The project showcases AI-generated interior design styles reflective of Taiwanese residential architecture, utilizing fine-tuned LGM models to ensure stylistic accuracy.

The process involves compiling categorized image datasets sourced from Taiwanese homes and applying LoRA fine-tuning with Kohya on Stable Diffusion 1.5 as the base model. Additionally, the design incorporates AI agents to propose relevant questions, with responses generated through fine-tuned LLMs, demonstrating a seamless integration of AI-driven workflows and technical precision.



f


Project Lead: Jimmy Wei-Chun Cheng
Synthetic Data Creation: Yi An Yang.
Technical Consultant: SHARE Lab, Karla Saldana Ochoa