Reproduce “One Shot, Any View: High-Fidelity Video Reshooting with Geometry Guidance and Visual References” 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
Kai Huang, Runze Xu, Hui Ren, Taiyi Wu, Xinyuan Li, Xun Gao, Quande Liu, Dong Tang and 3 more
One Shot Any View creates requested views of performances by reconstructing a time-varying 3D point cloud, rendering it along a camera path, and guiding a video generator with that rendering. Training combines real-camera and rendered data to limit geometry-defect copying. In 25 cases with under 50% point-video coverage, it had lower camera errors than Vista4D, but not in every subgroup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.