ANFluid: Animate Natural Fluid Photos base on Physics-Aware Simulation and Dual-Flow Texture Learning
Published in MM' 24, 2024
ANFluid introduces a new framework for generating realistic fluid animations from a single static image by combining physics-aware simulation (PAS) and dual-flow texture learning (DFTL). PAS ensures motion follows physical principles, while DFTL enhances texture prediction through innovative self-supervised techniques, improving animation quality without increasing model parameters. Experimental results show ANFluid outperforms existing methods in terms of physical consistency and content alignment, and user studies confirm its superior quality. The framework also supports interactive editing for dynamic content creation.
Recommended citation: Xiangcheng Zhai, Yingqi Jie, Xueguang Xie, Aimin Hao, Na Jiang, and Yang Gao. 2024. ANFluid':' Animate Natural Fluid Photos base on Physics-Aware Simulation and Dual-Flow Texture Learning. In Proceedings of the 32nd ACM International Conference on Multimedia (MM '24). Association for Computing Machinery, New York, NY, USA, 3323–3331. https://doi.org/10.1145/3664647.3680950