Reproduce “BirdDreams: Joint Perspective and BEV Video Generation for a Plausible, Reactive Driving World Model” 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
Junyoung Seo, Jaewoo Seo, Seungryong Kim, Laura Leal-Taixé, Xiuming Zhang, Seung Wook Kim
BirdDreams predicts camera video and a schematic bird’s-eye view together, with an optional ego path controlling the rollout. Its two streams exchange information, while the overhead representation provides a geometric condition rather than fixed forecasts of surrounding traffic. On PhysicalAI-AV, ego-path ADE is 3.83 versus 5.88 for OmniDreams; human ratings preferred BirdDreams in 57% of tested comparisons, not establishing safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.