Reproduce “IMPACT: Image-Conditioned Motion with Physics-Aware Collision and Trajectory Guidance for 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
Shiwei Yang, Yunlong Yu, Zhao Wang, Jungong Han
IMPACT guides image-to-video generation by estimating an initial physical state and simulating a sphere’s fall, impacts, diminishing rebounds, and settling. It uses the same rollout for sparse trajectory control and dense object-region guidance during diffusion sampling. On controlled sphere–plane scenes, human overall physical plausibility ratings were 3.61 versus 2.38 for sparse-only; this supports improved perceived realization, not exact physical fidelity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.