Reproduce “CompAdapt: Adaptable Composite Motion Modeling for Physics-Consistent Text-to-Video Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haoran Qin, Renlong Wu, Tianyu Huang, Yukang Ding, Hui Li, Wangmeng Zuo
CompAdapt extends NewtonGen to generate videos with composite motions, sequential stages, and elastic collisions from text-specified physical conditions. It parses prompts, composes neural Newtonian dynamics, and adapts a pretrained model using one observed trajectory. For projectile plus synchronous rotation, PIS is 0.9803 versus 0.8914 for retrained NewtonGen, while nonlinear coupled systems remain approximate.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.