Reproduce “WorldWander: Bridging Egocentric and Exocentric Worlds in 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
Quanjian Song, Yiren Song, Kelly Peng, Yuan Gao, Mike Zheng Shou
WorldWander learns bidirectional translation between egocentric and exocentric videos from synchronized examples, without camera poses or depth maps. It encodes conditions and targets jointly, keeps the condition clean while noising the target, and aligns tokens with attention and shared temporal positions. On real-world egocentric-to-exocentric FVD, it scores 374.7 versus 968.1 for fine-tuned ReCamMaster, evaluated per scene on 50 triplets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.