오토모티브 디자인을 위한 3D 모델링 생성 방법 연구 : 멀티뷰 이미지 입력 및 파츠 단위 변형을 중심으로

초록

With the rapid advancement of Image-to-3D generative AI, its potential application in early-stage automotive design has attracted increasing attention. This study analyzes the generation characteristics of commercial Image-to-3D services across different input-image representations and proposes two input-image design strategies for automotive 3D concept generation: multi-view input and part-level variation based on part-specific LoRA models with inpainting. An evaluation by ten automotive design experts showed that multi-view input supports early-stage review of vehicle form from the front, side, and rear views, overall proportions and volume, while part-level variation enables selective exploration of vehicle components without compromising the original body identity. However, limitations remain in cross-view consistency, transitions of part boundaries, reproducibility, and surface quality. These findings suggest that input-image design can serve as an effective approach for controlling the form of generated automotive 3D concept modeling and support part-level design exploration.

키워드