Qiwen Gu, Xin Zhou, Xiwu Chen, Zhongyang Zhu, Zhuo Zhang, Junqiao Zhao
FlexPart is a unified framework for image-to-3D part generation that supports variable-granularity geometric conditions. It integrates heterogeneous inputs and leverages spatial priors to improve geometric consistency. This framework reduces Part CD by 16.6% relative to existing methods.
Controllable 3D part generation is crucial for digital asset creation, and FlexPart's flexibility accommodates diverse user workflows. The framework's ability to integrate various inputs and improve geometric consistency makes it a valuable tool for applications like computer-aided design and video game development. This can enhance the efficiency and quality of 3D modeling tasks.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.